deaab59 Merge pull request 'refactor(agent): remove Ponytail mode' (#3) from refactor/remove-ponytail into main · zut

M README.md +4 -5


@@ -158,13 +158,13 @@ $ZUT_HOME/
158	158		└── logs/ # app log files
159	159		```
160	160
161		-	Drop a `SYSTEM.md` in `$ZUT_HOME` to replace the built-in identity and docs guidance for every run. Existing append addenda, including `AGENTS.md`, skills, and enabled Ponytail coding mode, remain layered on top. `--system-prompt` wins per invocation unless a selected replace-mode subagent profile supplies the child's identity. Delete the file to revert to the default identity.
	161	+	Drop a `SYSTEM.md` in `$ZUT_HOME` to replace the built-in identity and docs guidance for every run. Shared harness guidance and append addenda, including `AGENTS.md` and skills, remain layered on top. `--system-prompt` wins per invocation unless a selected replace-mode subagent profile supplies the child's identity. Delete the file to revert to the default identity.
162	162
163	163		Shared prompt guidance remains active with custom identities and across providers and models. It asks the agent to complete authorized, non-destructive work rather than stop at a plan, while respecting requests for explanation or review and preserving required approvals. Explicit user instructions take precedence over skill guidelines; skills cannot grant permissions or override system or developer constraints. Writing guidance favors concise paragraphs, useful lists, and direct language without canned transitions. These are prompt instructions, not changes to tool permissions or guarantees of model behavior.
164	164
165		-	Ponytail coding mode is enabled by default, including when an existing `config.json` does not contain a `ponytail_enabled` field. When disabled, its compact guidance is omitted from resolved system prompts; when enabled, the guidance tells the model to apply itself only to engineering work. The setting persists its explicit on/off choice in `$ZUT_HOME/config.json` and can be changed from `/settings`. It is advisory guidance only: it does not change tool permissions, confirmations, jail behavior, or sandboxing.
	165	+	Shared harness guidance asks agents to verify changes with checks appropriate to their scope and report which checks ran and what remains unverified. This guidance is included with custom identities and does not change tool permissions, confirmations, jail behavior, or sandboxing.
166	166
167		-	A selected subagent profile with `systemPromptMode: replace` supplies the child's base identity even when `--system-prompt` was provided for the parent run. Global append addenda, including enabled Ponytail guidance, still follow their normal inclusion rules.
	167	+	A selected subagent profile with `systemPromptMode: replace` supplies the child's base identity even when `--system-prompt` was provided for the parent run. Append-mode profiles retain shared harness guidance and global append addenda.
168	168
169	169		### HTTP proxy
170	170

@@ -354,7 +354,7 @@ Slash command names are case-insensitive in the TUI and messaging backends; argu
354	354		\| `/unjail` \| Allow tools to touch paths outside again. \|
355	355		\| `/reload-ext` \| Hot-reload all extensions (re-read manifests, respawn subprocesses, rebuild tool registry). \|
356	356		\| `/telegram` \| Connect, disconnect, or show status of the Telegram bridge (takes `connect` / `disconnect` / `status` as an optional argument; opens a picker without one). When connected, DMs from the paired user become prompts in the running session and the assistant's replies are mirrored back to Telegram. Alias: `/tg`. \|
357		-	\| `/settings` \| Change persistent settings, including web search, inline images, terminal alerts, AI terminal titles, proactive delegation, Ponytail coding mode, fast mode, main/sub-agent LSP access, and the auto-compact threshold. `/fast` and `/orchestrator` are shortcuts for their corresponding toggles. Saved to `$ZUT_HOME/config.json`; setting changes apply without a restart, while AI title generation waits for the first real prompt. \|
	357	+	\| `/settings` \| Change persistent settings, including web search, inline images, terminal alerts, AI terminal titles, proactive delegation, fast mode, main/sub-agent LSP access, and the auto-compact threshold. `/fast` and `/orchestrator` are shortcuts for their corresponding toggles. Saved to `$ZUT_HOME/config.json`; setting changes apply without a restart, while AI title generation waits for the first real prompt. \|
358	358		\| `/clear` \| Clear the chat transcript. \|
359	359		\| `/exit` \| Exit zut. \|
360	360

@@ -470,7 +470,6 @@ Opens a dialog with every persistent setting. `up`/`down` to navigate, `enter` o
470	470		- terminal alerts — emit a terminal bell when the main session stops with work that may need attention, or when an extension raises a structured alert. Enabled by default; changes apply immediately and persist as `terminal_alerts_enabled`. Terminal emulators may render the bell audibly, visually, or not at all.
471	471		- AI terminal titles — after the first real prompt of a fresh interactive session, make one small hidden request to the active model and set the terminal title to `zut: <title>`. The title is limited to 40 Unicode characters, persisted with the session, restored on resume, and never added to the conversation. Enabled by default; disable it to avoid the extra model request. The toggle applies immediately, but title generation still waits for the first real prompt and never starts from startup context, resumed history, or slash commands. Persists as `terminal_title_enabled`.
472	472		- proactive delegation — let the interactive primary delegate concrete, bounded sidecars while it keeps ownership of the critical path. Because the task and the workers it needs are unknown until the user states them, the primary routes at that point: the reply that starts the work names the parts it keeps on the critical path and the profile that owns each delegable part, and it states when nothing is worth delegating instead of forcing workers. Before spawning, it identifies useful local work and gives the child a non-overlapping question, responsibility, package, or file set. Immediate blockers and tightly coupled work stay local; after spawning, the worker owns its scope until completion and the primary continues only its preselected independent task. If a user request or active workflow explicitly delegates blocking work, the primary yields rather than reproducing it. Off by default. When off, the permitted canonical subagent tools and profile manifest remain available for user-requested delegation or when an active skill workflow requires it. The tools accept named profiles, reasoning, per-spawn fast-mode overrides, and shared/worktree isolation where applicable; lifecycle and result references are persisted. Completion and required-work state are host-driven through `[auto-subagents update]` and `[required-subagents update]` messages; required work stays asynchronous, persists its outcome, and prevents the terminal parent response until a successful completion. If launch-time restrictions prevent spawning, the interactive primary continues locally. Explicit headless `--orchestrate` remains a separate strict manager-only mode and reports unavailable delegation rather than implementing directly. Do not poll or inspect dashboards, metadata, event logs, or files solely to wait. See `/subagents` and [docs/subagents.md](docs/subagents.md) for details.
473		-	- Ponytail coding mode — include compact engineering guidance in each resolved system prompt; the guidance tells the model to apply it to coding, debugging, and review work rather than ordinary conversation. It favors understanding the real flow, small validated changes, reuse, and preserving safety checks. Enabled by default; changes apply to the next model call and persist as `ponytail_enabled` in `$ZUT_HOME/config.json`. It is included by interactive, print, stream, JSON, RPC, subagent, bot, SDK, and Zutfile agents.
474	473		- fast mode — request the provider's fast tier where supported. It currently uses OpenAI's Fast service tier for OpenAI, OpenAI Responses, and OpenAI Codex models. Off by default; toggle it with `/fast` or `/settings`; changes apply on the next model call and persist as `fast_mode`. When a model profile is active, the value is saved in that profile and restored when switching profiles. Other providers currently return an unsupported-provider error. Fast mode may cost more and depends on the selected model/account.
475	474		- lsp in main session — enable the built-in `lsp` tool and code diagnostics for the main agent. Enabled by default; changes persist as `lsp_enabled`.
476	475		- lsp in sub-agents — allow newly spawned background sub-agents to use the built-in `lsp` tool. Enabled by default; changes persist as `subagent_lsp_enabled` and apply when a child starts.

 └── logs/               # app log files
 ```
-Drop a `SYSTEM.md` in `$ZUT_HOME` to replace the built-in identity and docs guidance for every run. Existing append addenda, including `AGENTS.md`, skills, and enabled Ponytail coding mode, remain layered on top. `--system-prompt` wins per invocation unless a selected replace-mode subagent profile supplies the child's identity. Delete the file to revert to the default identity.
+Drop a `SYSTEM.md` in `$ZUT_HOME` to replace the built-in identity and docs guidance for every run. Shared harness guidance and append addenda, including `AGENTS.md` and skills, remain layered on top. `--system-prompt` wins per invocation unless a selected replace-mode subagent profile supplies the child's identity. Delete the file to revert to the default identity.
 Shared prompt guidance remains active with custom identities and across providers and models. It asks the agent to complete authorized, non-destructive work rather than stop at a plan, while respecting requests for explanation or review and preserving required approvals. Explicit user instructions take precedence over skill guidelines; skills cannot grant permissions or override system or developer constraints. Writing guidance favors concise paragraphs, useful lists, and direct language without canned transitions. These are prompt instructions, not changes to tool permissions or guarantees of model behavior.
-Ponytail coding mode is enabled by default, including when an existing `config.json` does not contain a `ponytail_enabled` field. When disabled, its compact guidance is omitted from resolved system prompts; when enabled, the guidance tells the model to apply itself only to engineering work. The setting persists its explicit on/off choice in `$ZUT_HOME/config.json` and can be changed from `/settings`. It is advisory guidance only: it does not change tool permissions, confirmations, jail behavior, or sandboxing.
+Shared harness guidance asks agents to verify changes with checks appropriate to their scope and report which checks ran and what remains unverified. This guidance is included with custom identities and does not change tool permissions, confirmations, jail behavior, or sandboxing.
-A selected subagent profile with `systemPromptMode: replace` supplies the child's base identity even when `--system-prompt` was provided for the parent run. Global append addenda, including enabled Ponytail guidance, still follow their normal inclusion rules.
+A selected subagent profile with `systemPromptMode: replace` supplies the child's base identity even when `--system-prompt` was provided for the parent run. Append-mode profiles retain shared harness guidance and global append addenda.
 ### HTTP proxy
 | `/unjail` | Allow tools to touch paths outside again. |
 | `/reload-ext` | Hot-reload all extensions (re-read manifests, respawn subprocesses, rebuild tool registry). |
 | `/telegram` | Connect, disconnect, or show status of the Telegram bridge (takes `connect` / `disconnect` / `status` as an optional argument; opens a picker without one). When connected, DMs from the paired user become prompts in the running session and the assistant's replies are mirrored back to Telegram. Alias: `/tg`. |
-| `/settings` | Change persistent settings, including web search, inline images, terminal alerts, AI terminal titles, proactive delegation, Ponytail coding mode, fast mode, main/sub-agent LSP access, and the auto-compact threshold. `/fast` and `/orchestrator` are shortcuts for their corresponding toggles. Saved to `$ZUT_HOME/config.json`; setting changes apply without a restart, while AI title generation waits for the first real prompt. |
+| `/settings` | Change persistent settings, including web search, inline images, terminal alerts, AI terminal titles, proactive delegation, fast mode, main/sub-agent LSP access, and the auto-compact threshold. `/fast` and `/orchestrator` are shortcuts for their corresponding toggles. Saved to `$ZUT_HOME/config.json`; setting changes apply without a restart, while AI title generation waits for the first real prompt. |
 | `/clear` | Clear the chat transcript. |
 | `/exit` | Exit zut. |
 - **terminal alerts** — emit a terminal bell when the main session stops with work that may need attention, or when an extension raises a structured alert. Enabled by default; changes apply immediately and persist as `terminal_alerts_enabled`. Terminal emulators may render the bell audibly, visually, or not at all.
 - **AI terminal titles** — after the first real prompt of a fresh interactive session, make one small hidden request to the active model and set the terminal title to `zut: <title>`. The title is limited to 40 Unicode characters, persisted with the session, restored on resume, and never added to the conversation. Enabled by default; disable it to avoid the extra model request. The toggle applies immediately, but title generation still waits for the first real prompt and never starts from startup context, resumed history, or slash commands. Persists as `terminal_title_enabled`.
 - **proactive delegation** — let the interactive primary delegate concrete, bounded sidecars while it keeps ownership of the critical path. Because the task and the workers it needs are unknown until the user states them, the primary routes at that point: the reply that starts the work names the parts it keeps on the critical path and the profile that owns each delegable part, and it states when nothing is worth delegating instead of forcing workers. Before spawning, it identifies useful local work and gives the child a non-overlapping question, responsibility, package, or file set. Immediate blockers and tightly coupled work stay local; after spawning, the worker owns its scope until completion and the primary continues only its preselected independent task. If a user request or active workflow explicitly delegates blocking work, the primary yields rather than reproducing it. Off by default. When off, the permitted canonical subagent tools and profile manifest remain available for user-requested delegation or when an active skill workflow requires it. The tools accept named profiles, reasoning, per-spawn fast-mode overrides, and shared/worktree isolation where applicable; lifecycle and result references are persisted. Completion and required-work state are host-driven through `[auto-subagents update]` and `[required-subagents update]` messages; required work stays asynchronous, persists its outcome, and prevents the terminal parent response until a successful completion. If launch-time restrictions prevent spawning, the interactive primary continues locally. Explicit headless `--orchestrate` remains a separate strict manager-only mode and reports unavailable delegation rather than implementing directly. Do not poll or inspect dashboards, metadata, event logs, or files solely to wait. See `/subagents` and [docs/subagents.md](docs/subagents.md) for details.
-- **Ponytail coding mode** — include compact engineering guidance in each resolved system prompt; the guidance tells the model to apply it to coding, debugging, and review work rather than ordinary conversation. It favors understanding the real flow, small validated changes, reuse, and preserving safety checks. Enabled by default; changes apply to the next model call and persist as `ponytail_enabled` in `$ZUT_HOME/config.json`. It is included by interactive, print, stream, JSON, RPC, subagent, bot, SDK, and Zutfile agents.
 - **fast mode** — request the provider's fast tier where supported. It currently uses OpenAI's Fast service tier for OpenAI, OpenAI Responses, and OpenAI Codex models. Off by default; toggle it with `/fast` or `/settings`; changes apply on the next model call and persist as `fast_mode`. When a model profile is active, the value is saved in that profile and restored when switching profiles. Other providers currently return an unsupported-provider error. Fast mode may cost more and depends on the selected model/account.
 - **lsp in main session** — enable the built-in `lsp` tool and code diagnostics for the main agent. Enabled by default; changes persist as `lsp_enabled`.
 - **lsp in sub-agents** — allow newly spawned background sub-agents to use the built-in `lsp` tool. Enabled by default; changes persist as `subagent_lsp_enabled` and apply when a child starts.

M docs/rpc.md +1 -1


@@ -25,7 +25,7 @@ You'll see one JSON object per line on stdout: a response acknowledging the prom
25	25
26	26		`zut rpc` accepts the same flags as the other modes: `--provider`, `--model`, `--cwd`, `--api-key`, `--base-url`, `--system-prompt`, `--append-system-prompt`, `--reasoning`, `--max-steps`, `--no-tools`, `--tools`, `--no-context-files`. Sessions are disabled by default in RPC mode — the embedding application owns persistence.
27	27
28		-	RPC agents use the same resolved system-prompt policy as other modes. Ponytail coding mode is enabled when `ponytail_enabled` is missing or `true` in `$ZUT_HOME/config.json`, and disabled when it is `false`. A custom `--system-prompt` replaces the built-in identity and docs guidance, while enabled append addenda remain present.
	28	+	RPC agents use the same resolved system-prompt policy as other modes. A custom `--system-prompt` replaces the built-in identity and docs guidance, while shared harness guidance and append addenda remain present. Shared guidance asks agents to verify changes with checks appropriate to their scope and report which checks ran and what remains unverified.
29	29
30	30		## Auth
31	31


@@ -25,7 +25,7 @@ You'll see one JSON object per line on stdout: a response acknowledging the prom
25			25
26		`zut rpc` accepts the same flags as the other modes: `--provider`, `--model`, `--cwd`, `--api-key`, `--base-url`, `--system-prompt`, `--append-system-prompt`, `--reasoning`, `--max-steps`, `--no-tools`, `--tools`, `--no-context-files`. Sessions are disabled by default in RPC mode — the embedding application owns persistence.	26		`zut rpc` accepts the same flags as the other modes: `--provider`, `--model`, `--cwd`, `--api-key`, `--base-url`, `--system-prompt`, `--append-system-prompt`, `--reasoning`, `--max-steps`, `--no-tools`, `--tools`, `--no-context-files`. Sessions are disabled by default in RPC mode — the embedding application owns persistence.
27			27
28	-	RPC agents use the same resolved system-prompt policy as other modes. Ponytail coding mode is enabled when `ponytail_enabled` is missing or `true` in `$ZUT_HOME/config.json`, and disabled when it is `false`. A custom `--system-prompt` replaces the built-in identity and docs guidance, while enabled append addenda remain present.	28	+	RPC agents use the same resolved system-prompt policy as other modes. A custom `--system-prompt` replaces the built-in identity and docs guidance, while shared harness guidance and append addenda remain present. Shared guidance asks agents to verify changes with checks appropriate to their scope and report which checks ran and what remains unverified.
29			29
30		## Auth	30		## Auth
31			31

M docs/zutfiles.md +1 -1


@@ -115,7 +115,7 @@ Set `replace_system_prompt` to `true` in the manifest to use `AGENT.md` as the r
115	115		}
116	116		```
117	117
118		-	Replacement is intended for fully specialized agents. It replaces zut's built-in identity, while global append addenda such as enabled Ponytail coding guidance, project context, and skills retain their normal inclusion rules. The default layering behavior is usually preferable because it retains zut's normal identity and tool-use guidance.
	118	+	Replacement is intended for fully specialized agents. It replaces zut's built-in identity, while shared harness guidance and global append addenda such as project context and skills retain their normal inclusion rules. The default layering behavior is usually preferable because it retains zut's normal identity and tool-use guidance.
119	119
120	120		## Bundled skills
121	121


@@ -115,7 +115,7 @@ Set `replace_system_prompt` to `true` in the manifest to use `AGENT.md` as the r
115		}	115		}
116		```	116		```
117			117
118	-	Replacement is intended for fully specialized agents. It replaces zut's built-in identity, while global append addenda such as enabled Ponytail coding guidance, project context, and skills retain their normal inclusion rules. The default layering behavior is usually preferable because it retains zut's normal identity and tool-use guidance.	118	+	Replacement is intended for fully specialized agents. It replaces zut's built-in identity, while shared harness guidance and global append addenda such as project context and skills retain their normal inclusion rules. The default layering behavior is usually preferable because it retains zut's normal identity and tool-use guidance.
119			119
120		## Bundled skills	120		## Bundled skills
121			121

M packages/agent/build.go +0 -5


@@ -962,11 +962,6 @@ func Resolve(args Args, requireCred bool) (Resolved, error) {
962	962		contextFiles = loadAgentsContext(args.CWD, ZutHome())
963	963		}
964	964		append_ := []string(nil)
965		-	if cfg.PonytailModeEnabled() {
966		-	if ponytailAddendum := PonytailSystemAddendum(); ponytailAddendum != "" {
967		-	append_ = append(append_, ponytailAddendum)
968		-	}
969		-	}
970	965		append_ = append(append_, args.AppendSystemPrompt...)
971	966		if agentsAddendum := formatAgentsContext(contextFiles); agentsAddendum != "" {
972	967		append_ = append(append_, agentsAddendum)


@@ -962,11 +962,6 @@ func Resolve(args Args, requireCred bool) (Resolved, error) {
962		contextFiles = loadAgentsContext(args.CWD, ZutHome())	962	contextFiles = loadAgentsContext(args.CWD, ZutHome())
963		}	963	}
964		append_ := []string(nil)	964	append_ := []string(nil)
965	-	if cfg.PonytailModeEnabled() {
966	-	if ponytailAddendum := PonytailSystemAddendum(); ponytailAddendum != "" {
967	-	append_ = append(append_, ponytailAddendum)
968	-	}
969	-	}
970		append_ = append(append_, args.AppendSystemPrompt...)	965	append_ = append(append_, args.AppendSystemPrompt...)
971		if agentsAddendum := formatAgentsContext(contextFiles); agentsAddendum != "" {	966	if agentsAddendum := formatAgentsContext(contextFiles); agentsAddendum != "" {
972		append_ = append(append_, agentsAddendum)	967	append_ = append(append_, agentsAddendum)

M packages/agent/cli.go +1 -11


@@ -2715,7 +2715,6 @@ func runInteractive(ctx context.Context, args Args, version string) (runErr erro
2715	2715		TerminalAlertsEnabled: initialCfg.TerminalAlertsEnabled,
2716	2716		TerminalTitleEnabled: initialCfg.TerminalTitleEnabled,
2717	2717		AutoSubagentsEnabled: initialCfg.AutoSubagentsEnabled,
2718		-	PonytailEnabled: initialCfg.PonytailEnabled,
2719	2718		WebSearchEnabled: initialCfg.WebSearchEnabled,
2720	2719		WebSearchToolAllowed: &webSearchToolAllowedForInvocation,
2721	2720		WebSearchInvocationOverride: webSearchInvocationOverride,

@@ -2809,16 +2808,7 @@ func runInteractive(ctx context.Context, args Args, version string) (runErr erro
2809	2808		setWebSearchAvailable(webSearchPolicy == subagents.WebSearchAllow)
2810	2809		return nil
2811	2810		},
2812		-	SetWebSearchAvailable: setWebSearchAvailable,
2813		-	RefreshPrompt: func() error {
2814		-	current := liveInteractiveAgent(iv, ag)
2815		-	webSearchPolicy, err := refreshAgentToolsAndPrompt(args, sharedSandbox, extToolAdapter, current, prepareInteractiveRegistryWithScheduler, iv)
2816		-	if err != nil {
2817		-	return err
2818		-	}
2819		-	setWebSearchAvailable(webSearchPolicy == subagents.WebSearchAllow)
2820		-	return nil
2821		-	},
	2811	+	SetWebSearchAvailable: setWebSearchAvailable,
2822	2812		AuthManager: mgr,
2823	2813		LlamaCPPConfig: ResolveLlamaCPPConfig,
2824	2814		RefreshLlamaCPPModels: RefreshLlamaCPPModels,

 		TerminalAlertsEnabled:             initialCfg.TerminalAlertsEnabled,
 		TerminalTitleEnabled:              initialCfg.TerminalTitleEnabled,
 		AutoSubagentsEnabled:              initialCfg.AutoSubagentsEnabled,
-		PonytailEnabled:                   initialCfg.PonytailEnabled,
 		WebSearchEnabled:                  initialCfg.WebSearchEnabled,
 		WebSearchToolAllowed:              &webSearchToolAllowedForInvocation,
 		WebSearchInvocationOverride:       webSearchInvocationOverride,
 			setWebSearchAvailable(webSearchPolicy == subagents.WebSearchAllow)
 			return nil
 		},
-		SetWebSearchAvailable: setWebSearchAvailable,
+		SetWebSearchAvailable:       setWebSearchAvailable,
-		RefreshPrompt: func() error {
-			current := liveInteractiveAgent(iv, ag)
-			webSearchPolicy, err := refreshAgentToolsAndPrompt(args, sharedSandbox, extToolAdapter, current, prepareInteractiveRegistryWithScheduler, iv)
-			if err != nil {
-				return err
-			}
-			setWebSearchAvailable(webSearchPolicy == subagents.WebSearchAllow)
-			return nil
-		},
 		AuthManager:                 mgr,
 		LlamaCPPConfig:              ResolveLlamaCPPConfig,
 		RefreshLlamaCPPModels:       RefreshLlamaCPPModels,

M packages/agent/config.go +0 -10


@@ -102,10 +102,6 @@ type Config struct {
102	102		// nil/missing means disabled. Toggle from /settings.
103	103		AutoSubagentsEnabled *bool `json:"auto_subagents_enabled,omitempty"`
104	104
105		-	// PonytailEnabled controls the compact coding-guidance addendum. On by
106		-	// default; nil/missing means enabled. Toggle from /settings.
107		-	PonytailEnabled *bool `json:"ponytail_enabled,omitempty"`
108		-
109	105		// WebSearchEnabled controls the built-in web_search capability for
110	106		// normal CLI sessions. nil/missing means enabled so existing config
111	107		// files retain the default-on behavior.

@@ -274,12 +270,6 @@ func (c Config) LSPEnabledFor(subagent bool) bool {
274	270		return value == nil \|\| *value
275	271		}
276	272
277		-	// PonytailModeEnabled reports the default-on coding-guidance preference
278		-	// without requiring a generated config entry.
279		-	func (c Config) PonytailModeEnabled() bool {
280		-	return c.PonytailEnabled == nil \|\| *c.PonytailEnabled
281		-	}
282		-
283	273		// WebSearchEnabledForCLI reports the default-on web-search preference
284	274		// without requiring a generated config entry.
285	275		func (c Config) WebSearchEnabledForCLI() bool {


@@ -102,10 +102,6 @@ type Config struct {
102		// nil/missing means disabled. Toggle from /settings.	102	// nil/missing means disabled. Toggle from /settings.
103		AutoSubagentsEnabled *bool `json:"auto_subagents_enabled,omitempty"`	103	AutoSubagentsEnabled *bool `json:"auto_subagents_enabled,omitempty"`
104			104
105	-	// PonytailEnabled controls the compact coding-guidance addendum. On by
106	-	// default; nil/missing means enabled. Toggle from /settings.
107	-	PonytailEnabled *bool `json:"ponytail_enabled,omitempty"`
108	-
109		// WebSearchEnabled controls the built-in web_search capability for	105	// WebSearchEnabled controls the built-in web_search capability for
110		// normal CLI sessions. nil/missing means enabled so existing config	106	// normal CLI sessions. nil/missing means enabled so existing config
111		// files retain the default-on behavior.	107	// files retain the default-on behavior.

@@ -274,12 +270,6 @@ func (c Config) LSPEnabledFor(subagent bool) bool {
274		return value == nil \|\| *value	270	return value == nil \|\| *value
275		}	271	}
276			272
277	-	// PonytailModeEnabled reports the default-on coding-guidance preference
278	-	// without requiring a generated config entry.
279	-	func (c Config) PonytailModeEnabled() bool {
280	-	return c.PonytailEnabled == nil \|\| *c.PonytailEnabled
281	-	}
282	-
283		// WebSearchEnabledForCLI reports the default-on web-search preference	273	// WebSearchEnabledForCLI reports the default-on web-search preference
284		// without requiring a generated config entry.	274	// without requiring a generated config entry.
285		func (c Config) WebSearchEnabledForCLI() bool {	275	func (c Config) WebSearchEnabledForCLI() bool {

M packages/agent/modes/interactive.go +1 -14


@@ -69,10 +69,6 @@ type InteractiveConfig struct {
69	69		// re-reading config.json on every open.
70	70		AutoSubagentsEnabled *bool
71	71
72		-	// PonytailEnabled mirrors the persisted default-on coding-guidance
73		-	// setting at startup. nil/missing means enabled.
74		-	PonytailEnabled *bool
75		-
76	72		// WebSearchEnabled mirrors the persisted default-on web-search setting.
77	73		// nil/missing means enabled for normal interactive sessions.
78	74		WebSearchEnabled *bool

@@ -232,10 +228,6 @@ type InteractiveConfig struct {
232	228		// web-search guard and resident child policy.
233	229		SetWebSearchAvailable func(bool)
234	230
235		-	// RefreshPrompt re-resolves the complete live prompt and tool registry
236		-	// after a prompt-affecting setting changes. Optional for embedders.
237		-	RefreshPrompt func() error
238		-
239	231		// AutoSubmitInitial, when true, auto-submits InitialInput after
240	232		// StartupPre completes (or immediately when StartupPre is empty).
241	233		// When false, InitialInput only pre-fills the editor.

@@ -586,10 +578,6 @@ type fastModeSettingsStore interface {
586	578		SetFastMode(enabled bool) error
587	579		}
588	580
589		-	type ponytailSettingsStore interface {
590		-	SetPonytailEnabled(enabled bool) error
591		-	}
592		-
593	581		type webSearchSettingsStore interface {
594	582		SetWebSearchEnabled(enabled bool) error
595	583		}

@@ -1994,8 +1982,7 @@ const modalBackdropDimPercent = 50
1994	1982		// hold i.mu.
1995	1983
1996	1984		// autoCompactContinuationPrompt is an interactive-only compact-handoff
1997		-	// instruction. Its completion criteria must stay separate from Ponytail so
1998		-	// toggling the optional coding-style addendum never changes recovery behavior.
	1985	+	// instruction defining recovery completion criteria.
1999	1986		const autoCompactContinuationPrompt = `Context was compacted while work was in progress. Continue the user's most recent active request now; do not wait for them to type "continue". Inspect the context summary and kept recent messages, treating active constraints and preferences there as still in force, including delegation/subagent instructions. If a newer request supersedes earlier plans in the summary, follow the newer request. A progress report, plan update, or statement of future work is not completion: if required work remains, take the next concrete action now. Only finish when the request is actually complete or when a specific user decision is required. If nothing remains, give a brief truthful completion; do not invent work or force a tool call.`
2000	1987
2001	1988		type compactHandoffResume uint8

 	// re-reading config.json on every open.
 	AutoSubagentsEnabled *bool
-	// PonytailEnabled mirrors the persisted default-on coding-guidance
-	// setting at startup. nil/missing means enabled.
-	PonytailEnabled *bool
 	// WebSearchEnabled mirrors the persisted default-on web-search setting.
 	// nil/missing means enabled for normal interactive sessions.
 	WebSearchEnabled *bool
 	// web-search guard and resident child policy.
 	SetWebSearchAvailable func(bool)
-	// RefreshPrompt re-resolves the complete live prompt and tool registry
-	// after a prompt-affecting setting changes. Optional for embedders.
-	RefreshPrompt func() error
 	// AutoSubmitInitial, when true, auto-submits InitialInput after
 	// StartupPre completes (or immediately when StartupPre is empty).
 	// When false, InitialInput only pre-fills the editor.
 	SetFastMode(enabled bool) error
 }
-type ponytailSettingsStore interface {
-	SetPonytailEnabled(enabled bool) error
-}
 type webSearchSettingsStore interface {
 	SetWebSearchEnabled(enabled bool) error
 }
 // hold i.mu.
 // autoCompactContinuationPrompt is an interactive-only compact-handoff
-// instruction. Its completion criteria must stay separate from Ponytail so
+// instruction defining recovery completion criteria.
-// toggling the optional coding-style addendum never changes recovery behavior.
 const autoCompactContinuationPrompt = `Context was compacted while work was in progress. Continue the user's most recent active request now; do not wait for them to type "continue". Inspect the context summary and kept recent messages, treating active constraints and preferences there as still in force, including delegation/subagent instructions. If a newer request supersedes earlier plans in the summary, follow the newer request. A progress report, plan update, or statement of future work is not completion: if required work remains, take the next concrete action now. Only finish when the request is actually complete or when a specific user decision is required. If nothing remains, give a brief truthful completion; do not invent work or force a tool call.`
 type compactHandoffResume uint8

M packages/agent/modes/interactive_settings.go +0 -72


@@ -158,13 +158,6 @@ func (i *Interactive) openSettingsDialog() {
158	158		autoSubagents = false
159	159		}
160	160
161		-	ponytailEnabled := i.ponytailEnabled()
162		-	ponytailDisabled := !i.ponytailSettingsAvailable()
163		-	ponytailHint := ""
164		-	if ponytailDisabled {
165		-	ponytailHint = "requires persistent settings and live prompt refresh support"
166		-	}
167		-
168	161		webSearchEnabled := i.webSearchEffectiveEnabled()
169	162		webSearchDisabled := !i.webSearchSettingsAvailable()
170	163		webSearchHint := i.webSearchUnavailableHint()

@@ -296,14 +289,6 @@ func (i *Interactive) openSettingsDialog() {
296	289		disabled: autoSubagentsDisabled,
297	290		hint: autoSubagentsHint,
298	291		},
299		-	{
300		-	key: "ponytail_enabled",
301		-	label: "ponytail coding mode",
302		-	desc: "apply compact coding guidance that favors small, validated, maintainable changes",
303		-	value: ponytailEnabled,
304		-	disabled: ponytailDisabled,
305		-	hint: ponytailHint,
306		-	},
307	292		{
308	293		key: "web_search_enabled",
309	294		label: "web search",

@@ -645,13 +630,6 @@ func (i *Interactive) resetSettingsToggle(key string, value bool) {
645	630		}
646	631		}
647	632		}
648		-	func (i *Interactive) ponytailEnabled() bool {
649		-	return i.cfg.PonytailEnabled == nil \|\| *i.cfg.PonytailEnabled
650		-	}
651		-	func (i *Interactive) ponytailSettingsAvailable() bool {
652		-	_, persistent := i.cfg.SettingsStore.(ponytailSettingsStore)
653		-	return persistent && i.cfg.RefreshPrompt != nil
654		-	}
655	633		func (i *Interactive) webSearchEnabled() bool {
656	634		return i.cfg.WebSearchEnabled == nil \|\| *i.cfg.WebSearchEnabled
657	635		}

@@ -847,56 +825,6 @@ func (i *Interactive) applySettingToggle(key string, value bool) {
847	825		} else {
848	826		apply()
849	827		}
850		-	case "ponytail_enabled":
851		-	previous := i.ponytailEnabled()
852		-	store, available := i.cfg.SettingsStore.(ponytailSettingsStore)
853		-	if !available \|\| i.cfg.RefreshPrompt == nil {
854		-	i.resetSettingsToggle(key, previous)
855		-	i.mu.Lock()
856		-	i.statusOK = ""
857		-	i.statusErr = "ponytail coding mode unavailable: persistent settings and live prompt refresh are required"
858		-	i.mu.Unlock()
859		-	return
860		-	}
861		-	if err := store.SetPonytailEnabled(value); err != nil {
862		-	i.resetSettingsToggle(key, previous)
863		-	i.mu.Lock()
864		-	i.statusOK = ""
865		-	i.statusErr = "settings: " + err.Error()
866		-	i.mu.Unlock()
867		-	return
868		-	}
869		-	val := value
870		-	i.cfg.PonytailEnabled = &val
871		-	if err := i.cfg.RefreshPrompt(); err != nil {
872		-	errMsg := "settings: ponytail prompt refresh: " + err.Error()
873		-	if rollbackErr := store.SetPonytailEnabled(previous); rollbackErr != nil {
874		-	// The first write succeeded, so keep the in-memory setting at
875		-	// the value that remains durable instead of claiming that the
876		-	// rollback took effect.
877		-	i.resetSettingsToggle(key, value)
878		-	errMsg += "; rollback persistence: " + rollbackErr.Error()
879		-	if reconcileErr := i.cfg.RefreshPrompt(); reconcileErr != nil {
880		-	errMsg += "; durable-state refresh: " + reconcileErr.Error()
881		-	}
882		-	} else {
883		-	previousVal := previous
884		-	i.cfg.PonytailEnabled = &previousVal
885		-	i.resetSettingsToggle(key, previous)
886		-	if refreshErr := i.cfg.RefreshPrompt(); refreshErr != nil {
887		-	errMsg += "; rollback refresh: " + refreshErr.Error()
888		-	}
889		-	}
890		-	i.mu.Lock()
891		-	i.statusOK = ""
892		-	i.statusErr = errMsg
893		-	i.mu.Unlock()
894		-	return
895		-	}
896		-	i.mu.Lock()
897		-	i.statusOK = "ponytail coding mode " + onOff(value)
898		-	i.statusErr = ""
899		-	i.mu.Unlock()
900	828		case "auto_subagents_enabled":
901	829		if value && !i.autoSubagentsAvailable() {
902	830		i.mu.Lock()


@@ -158,13 +158,6 @@ func (i *Interactive) openSettingsDialog() {
158		autoSubagents = false	158	autoSubagents = false
159		}	159	}
160			160
161	-	ponytailEnabled := i.ponytailEnabled()
162	-	ponytailDisabled := !i.ponytailSettingsAvailable()
163	-	ponytailHint := ""
164	-	if ponytailDisabled {
165	-	ponytailHint = "requires persistent settings and live prompt refresh support"
166	-	}
167	-
168		webSearchEnabled := i.webSearchEffectiveEnabled()	161	webSearchEnabled := i.webSearchEffectiveEnabled()
169		webSearchDisabled := !i.webSearchSettingsAvailable()	162	webSearchDisabled := !i.webSearchSettingsAvailable()
170		webSearchHint := i.webSearchUnavailableHint()	163	webSearchHint := i.webSearchUnavailableHint()

@@ -296,14 +289,6 @@ func (i *Interactive) openSettingsDialog() {
296		disabled: autoSubagentsDisabled,	289	disabled: autoSubagentsDisabled,
297		hint: autoSubagentsHint,	290	hint: autoSubagentsHint,
298		},	291	},
299	-	{
300	-	key: "ponytail_enabled",
301	-	label: "ponytail coding mode",
302	-	desc: "apply compact coding guidance that favors small, validated, maintainable changes",
303	-	value: ponytailEnabled,
304	-	disabled: ponytailDisabled,
305	-	hint: ponytailHint,
306	-	},
307		{	292	{
308		key: "web_search_enabled",	293	key: "web_search_enabled",
309		label: "web search",	294	label: "web search",

@@ -645,13 +630,6 @@ func (i *Interactive) resetSettingsToggle(key string, value bool) {
645		}	630	}
646		}	631	}
647		}	632	}
648	-	func (i *Interactive) ponytailEnabled() bool {
649	-	return i.cfg.PonytailEnabled == nil \|\| *i.cfg.PonytailEnabled
650	-	}
651	-	func (i *Interactive) ponytailSettingsAvailable() bool {
652	-	_, persistent := i.cfg.SettingsStore.(ponytailSettingsStore)
653	-	return persistent && i.cfg.RefreshPrompt != nil
654	-	}
655		func (i *Interactive) webSearchEnabled() bool {	633	func (i *Interactive) webSearchEnabled() bool {
656		return i.cfg.WebSearchEnabled == nil \|\| *i.cfg.WebSearchEnabled	634	return i.cfg.WebSearchEnabled == nil \|\| *i.cfg.WebSearchEnabled
657		}	635	}

@@ -847,56 +825,6 @@ func (i *Interactive) applySettingToggle(key string, value bool) {
847		} else {	825	} else {
848		apply()	826	apply()
849		}	827	}
850	-	case "ponytail_enabled":
851	-	previous := i.ponytailEnabled()
852	-	store, available := i.cfg.SettingsStore.(ponytailSettingsStore)
853	-	if !available \|\| i.cfg.RefreshPrompt == nil {
854	-	i.resetSettingsToggle(key, previous)
855	-	i.mu.Lock()
856	-	i.statusOK = ""
857	-	i.statusErr = "ponytail coding mode unavailable: persistent settings and live prompt refresh are required"
858	-	i.mu.Unlock()
859	-	return
860	-	}
861	-	if err := store.SetPonytailEnabled(value); err != nil {
862	-	i.resetSettingsToggle(key, previous)
863	-	i.mu.Lock()
864	-	i.statusOK = ""
865	-	i.statusErr = "settings: " + err.Error()
866	-	i.mu.Unlock()
867	-	return
868	-	}
869	-	val := value
870	-	i.cfg.PonytailEnabled = &val
871	-	if err := i.cfg.RefreshPrompt(); err != nil {
872	-	errMsg := "settings: ponytail prompt refresh: " + err.Error()
873	-	if rollbackErr := store.SetPonytailEnabled(previous); rollbackErr != nil {
874	-	// The first write succeeded, so keep the in-memory setting at
875	-	// the value that remains durable instead of claiming that the
876	-	// rollback took effect.
877	-	i.resetSettingsToggle(key, value)
878	-	errMsg += "; rollback persistence: " + rollbackErr.Error()
879	-	if reconcileErr := i.cfg.RefreshPrompt(); reconcileErr != nil {
880	-	errMsg += "; durable-state refresh: " + reconcileErr.Error()
881	-	}
882	-	} else {
883	-	previousVal := previous
884	-	i.cfg.PonytailEnabled = &previousVal
885	-	i.resetSettingsToggle(key, previous)
886	-	if refreshErr := i.cfg.RefreshPrompt(); refreshErr != nil {
887	-	errMsg += "; rollback refresh: " + refreshErr.Error()
888	-	}
889	-	}
890	-	i.mu.Lock()
891	-	i.statusOK = ""
892	-	i.statusErr = errMsg
893	-	i.mu.Unlock()
894	-	return
895	-	}
896	-	i.mu.Lock()
897	-	i.statusOK = "ponytail coding mode " + onOff(value)
898	-	i.statusErr = ""
899	-	i.mu.Unlock()
900		case "auto_subagents_enabled":	828	case "auto_subagents_enabled":
901		if value && !i.autoSubagentsAvailable() {	829	if value && !i.autoSubagentsAvailable() {
902		i.mu.Lock()	830	i.mu.Lock()

D packages/agent/modes/ponytail_setting_test.go +0 -230

@@ -1,230 +0,0 @@
1	-	package modes
2	-
3	-	import (
4	-	"errors"
5	-	"strings"
6	-	"testing"
7	-
8	-	"github.com/bnema/zut/packages/core"
9	-	)
10	-
11	-	type recordingPonytailSettingsStore struct {
12	-	SettingsStore
13	-	values []bool
14	-	calls int
15	-	err error
16	-	failures map[int]error
17	-	}
18	-
19	-	func (s *recordingPonytailSettingsStore) SetPonytailEnabled(enabled bool) error {
20	-	s.calls++
21	-	if s.err != nil {
22	-	return s.err
23	-	}
24	-	if err := s.failures[s.calls]; err != nil {
25	-	return err
26	-	}
27	-	s.values = append(s.values, enabled)
28	-	return nil
29	-	}
30	-
31	-	func TestSettingsDialogOffersPonytailEnabledByDefault(t *testing.T) {
32	-	store := &recordingPonytailSettingsStore{}
33	-	interactive := NewInteractive(InteractiveConfig{
34	-	SettingsStore: store,
35	-	RefreshPrompt: func() error { return nil },
36	-	})
37	-	interactive.openSettingsDialog()
38	-
39	-	item := findSettingsItem(interactive.settingsDialog.items, "ponytail_enabled")
40	-	if item == nil {
41	-	t.Fatal("settings dialog is missing Ponytail mode")
42	-	}
43	-	if !item.value {
44	-	t.Fatal("Ponytail mode should be enabled when the config value is absent")
45	-	}
46	-	if item.disabled {
47	-	t.Fatal("Ponytail mode should be available with persistence and refresh capabilities")
48	-	}
49	-	}
50	-
51	-	func TestSettingsDialogDisablesPonytailWithoutCapabilities(t *testing.T) {
52	-	cases := []struct {
53	-	name string
54	-	cfg InteractiveConfig
55	-	}{
56	-	{name: "neither capability", cfg: InteractiveConfig{}},
57	-	{name: "persistence only", cfg: InteractiveConfig{SettingsStore: &recordingPonytailSettingsStore{}}},
58	-	{name: "refresh only", cfg: InteractiveConfig{RefreshPrompt: func() error { return nil }}},
59	-	}
60	-	for _, tc := range cases {
61	-	t.Run(tc.name, func(t *testing.T) {
62	-	interactive := NewInteractive(tc.cfg)
63	-	interactive.openSettingsDialog()
64	-
65	-	item := findSettingsItem(interactive.settingsDialog.items, "ponytail_enabled")
66	-	if item == nil {
67	-	t.Fatal("settings dialog is missing Ponytail mode")
68	-	}
69	-	if !item.disabled {
70	-	t.Fatal("Ponytail mode should be disabled without both capabilities")
71	-	}
72	-	if item.hint == "" {
73	-	t.Fatal("disabled Ponytail mode should explain why it is unavailable")
74	-	}
75	-	})
76	-	}
77	-	}
78	-
79	-	func TestApplyPonytailToggleRefreshesAndPersists(t *testing.T) {
80	-	store := &recordingPonytailSettingsStore{}
81	-	refreshCalls := 0
82	-	interactive := NewInteractive(InteractiveConfig{
83	-	SettingsStore: store,
84	-	RefreshPrompt: func() error {
85	-	refreshCalls++
86	-	return nil
87	-	},
88	-	})
89	-	interactive.rend = nil
90	-
91	-	interactive.applySettingToggle("ponytail_enabled", false)
92	-	if interactive.ponytailEnabled() {
93	-	t.Fatal("Ponytail mode remained enabled after disabling it")
94	-	}
95	-	interactive.applySettingToggle("ponytail_enabled", true)
96	-	if !interactive.ponytailEnabled() {
97	-	t.Fatal("Ponytail mode remained disabled after enabling it")
98	-	}
99	-	if refreshCalls != 2 {
100	-	t.Fatalf("refresh calls = %d, want 2", refreshCalls)
101	-	}
102	-	if len(store.values) != 2 \|\| store.values[0] \|\| !store.values[1] {
103	-	t.Fatalf("persisted values = %v, want [false true]", store.values)
104	-	}
105	-	if interactive.statusErr != "" {
106	-	t.Fatalf("unexpected status error: %q", interactive.statusErr)
107	-	}
108	-	}
109	-
110	-	func TestApplyPonytailToggleRollsBackWhenRefreshFails(t *testing.T) {
111	-	store := &recordingPonytailSettingsStore{}
112	-	refreshCalls := 0
113	-	interactive := NewInteractive(InteractiveConfig{
114	-	SettingsStore: store,
115	-	RefreshPrompt: func() error {
116	-	refreshCalls++
117	-	if refreshCalls == 1 {
118	-	return errors.New("refresh failed")
119	-	}
120	-	return nil
121	-	},
122	-	})
123	-	interactive.rend = nil
124	-
125	-	interactive.applySettingToggle("ponytail_enabled", false)
126	-	if !interactive.ponytailEnabled() {
127	-	t.Fatal("Ponytail mode stayed disabled after refresh rollback")
128	-	}
129	-	if refreshCalls != 2 {
130	-	t.Fatalf("refresh calls = %d, want initial refresh and rollback refresh", refreshCalls)
131	-	}
132	-	if len(store.values) != 2 \|\| store.values[0] \|\| !store.values[1] {
133	-	t.Fatalf("persisted rollback values = %v, want [false true]", store.values)
134	-	}
135	-	if !strings.Contains(interactive.statusErr, "refresh failed") {
136	-	t.Fatalf("status error = %q, want refresh failure", interactive.statusErr)
137	-	}
138	-	}
139	-
140	-	func TestApplyPonytailToggleKeepsDurableStateWhenRollbackPersistenceFails(t *testing.T) {
141	-	store := &recordingPonytailSettingsStore{
142	-	failures: map[int]error{2: errors.New("rollback disk full")},
143	-	}
144	-	refreshCalls := 0
145	-	interactive := NewInteractive(InteractiveConfig{
146	-	SettingsStore: store,
147	-	RefreshPrompt: func() error {
148	-	refreshCalls++
149	-	if refreshCalls == 1 {
150	-	return errors.New("refresh failed")
151	-	}
152	-	return nil
153	-	},
154	-	})
155	-	interactive.rend = nil
156	-
157	-	interactive.applySettingToggle("ponytail_enabled", false)
158	-	if interactive.ponytailEnabled() {
159	-	t.Fatal("in-memory setting reverted despite rollback persistence failure")
160	-	}
161	-	if len(store.values) != 1 \|\| store.values[0] {
162	-	t.Fatalf("durable writes = %v, want only the successful disable", store.values)
163	-	}
164	-	if refreshCalls != 2 {
165	-	t.Fatalf("refresh calls = %d, want failed refresh plus durable-state reconciliation", refreshCalls)
166	-	}
167	-	if !strings.Contains(interactive.statusErr, "rollback disk full") {
168	-	t.Fatalf("status error = %q, want rollback failure", interactive.statusErr)
169	-	}
170	-	}
171	-
172	-	func TestApplyPonytailToggleKeepsStateWhenPersistenceFails(t *testing.T) {
173	-	store := &recordingPonytailSettingsStore{err: errors.New("disk full")}
174	-	refreshCalls := 0
175	-	interactive := NewInteractive(InteractiveConfig{
176	-	SettingsStore: store,
177	-	RefreshPrompt: func() error {
178	-	refreshCalls++
179	-	return nil
180	-	},
181	-	})
182	-	interactive.rend = nil
183	-
184	-	interactive.applySettingToggle("ponytail_enabled", false)
185	-	if !interactive.ponytailEnabled() {
186	-	t.Fatal("Ponytail mode changed despite persistence failure")
187	-	}
188	-	if refreshCalls != 0 {
189	-	t.Fatalf("refresh calls = %d, want none", refreshCalls)
190	-	}
191	-	if !strings.Contains(interactive.statusErr, "disk full") {
192	-	t.Fatalf("status error = %q, want persistence failure", interactive.statusErr)
193	-	}
194	-	}
195	-
196	-	func TestApplyPonytailToggleUpdatesLiveAgentPrompt(t *testing.T) {
197	-	const ponytailBlock = "[ponytail test block]"
198	-	store := &recordingPonytailSettingsStore{}
199	-	ag := core.NewAgent(nil, "model", "base\n\n"+ponytailBlock, nil)
200	-	var interactive *Interactive
201	-	refreshCalls := 0
202	-	interactive = NewInteractive(InteractiveConfig{
203	-	Agent: ag,
204	-	SettingsStore: store,
205	-	RefreshPrompt: func() error {
206	-	refreshCalls++
207	-	system := "base"
208	-	if interactive.ponytailEnabled() {
209	-	system += "\n\n" + ponytailBlock
210	-	}
211	-	ag.SetSystemPrompt(system)
212	-	return nil
213	-	},
214	-	})
215	-	interactive.rend = nil
216	-
217	-	interactive.applySettingToggle("ponytail_enabled", false)
218	-	system, _ := ag.PromptConfig()
219	-	if strings.Contains(system, ponytailBlock) {
220	-	t.Fatalf("live prompt still contains disabled Ponytail block: %q", system)
221	-	}
222	-	interactive.applySettingToggle("ponytail_enabled", true)
223	-	system, _ = ag.PromptConfig()
224	-	if strings.Count(system, ponytailBlock) != 1 {
225	-	t.Fatalf("live prompt block count = %d, want 1: %q", strings.Count(system, ponytailBlock), system)
226	-	}
227	-	if refreshCalls != 2 {
228	-	t.Fatalf("refresh calls = %d, want 2", refreshCalls)
229	-	}
230	-	}

@@ -1,230 +0,0 @@
1	-	package modes
2	-
3	-	import (
4	-	"errors"
5	-	"strings"
6	-	"testing"
7	-
8	-	"github.com/bnema/zut/packages/core"
9	-	)
10	-
11	-	type recordingPonytailSettingsStore struct {
12	-	SettingsStore
13	-	values []bool
14	-	calls int
15	-	err error
16	-	failures map[int]error
17	-	}
18	-
19	-	func (s *recordingPonytailSettingsStore) SetPonytailEnabled(enabled bool) error {
20	-	s.calls++
21	-	if s.err != nil {
22	-	return s.err
23	-	}
24	-	if err := s.failures[s.calls]; err != nil {
25	-	return err
26	-	}
27	-	s.values = append(s.values, enabled)
28	-	return nil
29	-	}
30	-
31	-	func TestSettingsDialogOffersPonytailEnabledByDefault(t *testing.T) {
32	-	store := &recordingPonytailSettingsStore{}
33	-	interactive := NewInteractive(InteractiveConfig{
34	-	SettingsStore: store,
35	-	RefreshPrompt: func() error { return nil },
36	-	})
37	-	interactive.openSettingsDialog()
38	-
39	-	item := findSettingsItem(interactive.settingsDialog.items, "ponytail_enabled")
40	-	if item == nil {
41	-	t.Fatal("settings dialog is missing Ponytail mode")
42	-	}
43	-	if !item.value {
44	-	t.Fatal("Ponytail mode should be enabled when the config value is absent")
45	-	}
46	-	if item.disabled {
47	-	t.Fatal("Ponytail mode should be available with persistence and refresh capabilities")
48	-	}
49	-	}
50	-
51	-	func TestSettingsDialogDisablesPonytailWithoutCapabilities(t *testing.T) {
52	-	cases := []struct {
53	-	name string
54	-	cfg InteractiveConfig
55	-	}{
56	-	{name: "neither capability", cfg: InteractiveConfig{}},
57	-	{name: "persistence only", cfg: InteractiveConfig{SettingsStore: &recordingPonytailSettingsStore{}}},
58	-	{name: "refresh only", cfg: InteractiveConfig{RefreshPrompt: func() error { return nil }}},
59	-	}
60	-	for _, tc := range cases {
61	-	t.Run(tc.name, func(t *testing.T) {
62	-	interactive := NewInteractive(tc.cfg)
63	-	interactive.openSettingsDialog()
64	-
65	-	item := findSettingsItem(interactive.settingsDialog.items, "ponytail_enabled")
66	-	if item == nil {
67	-	t.Fatal("settings dialog is missing Ponytail mode")
68	-	}
69	-	if !item.disabled {
70	-	t.Fatal("Ponytail mode should be disabled without both capabilities")
71	-	}
72	-	if item.hint == "" {
73	-	t.Fatal("disabled Ponytail mode should explain why it is unavailable")
74	-	}
75	-	})
76	-	}
77	-	}
78	-
79	-	func TestApplyPonytailToggleRefreshesAndPersists(t *testing.T) {
80	-	store := &recordingPonytailSettingsStore{}
81	-	refreshCalls := 0
82	-	interactive := NewInteractive(InteractiveConfig{
83	-	SettingsStore: store,
84	-	RefreshPrompt: func() error {
85	-	refreshCalls++
86	-	return nil
87	-	},
88	-	})
89	-	interactive.rend = nil
90	-
91	-	interactive.applySettingToggle("ponytail_enabled", false)
92	-	if interactive.ponytailEnabled() {
93	-	t.Fatal("Ponytail mode remained enabled after disabling it")
94	-	}
95	-	interactive.applySettingToggle("ponytail_enabled", true)
96	-	if !interactive.ponytailEnabled() {
97	-	t.Fatal("Ponytail mode remained disabled after enabling it")
98	-	}
99	-	if refreshCalls != 2 {
100	-	t.Fatalf("refresh calls = %d, want 2", refreshCalls)
101	-	}
102	-	if len(store.values) != 2 \|\| store.values[0] \|\| !store.values[1] {
103	-	t.Fatalf("persisted values = %v, want [false true]", store.values)
104	-	}
105	-	if interactive.statusErr != "" {
106	-	t.Fatalf("unexpected status error: %q", interactive.statusErr)
107	-	}
108	-	}
109	-
110	-	func TestApplyPonytailToggleRollsBackWhenRefreshFails(t *testing.T) {
111	-	store := &recordingPonytailSettingsStore{}
112	-	refreshCalls := 0
113	-	interactive := NewInteractive(InteractiveConfig{
114	-	SettingsStore: store,
115	-	RefreshPrompt: func() error {
116	-	refreshCalls++
117	-	if refreshCalls == 1 {
118	-	return errors.New("refresh failed")
119	-	}
120	-	return nil
121	-	},
122	-	})
123	-	interactive.rend = nil
124	-
125	-	interactive.applySettingToggle("ponytail_enabled", false)
126	-	if !interactive.ponytailEnabled() {
127	-	t.Fatal("Ponytail mode stayed disabled after refresh rollback")
128	-	}
129	-	if refreshCalls != 2 {
130	-	t.Fatalf("refresh calls = %d, want initial refresh and rollback refresh", refreshCalls)
131	-	}
132	-	if len(store.values) != 2 \|\| store.values[0] \|\| !store.values[1] {
133	-	t.Fatalf("persisted rollback values = %v, want [false true]", store.values)
134	-	}
135	-	if !strings.Contains(interactive.statusErr, "refresh failed") {
136	-	t.Fatalf("status error = %q, want refresh failure", interactive.statusErr)
137	-	}
138	-	}
139	-
140	-	func TestApplyPonytailToggleKeepsDurableStateWhenRollbackPersistenceFails(t *testing.T) {
141	-	store := &recordingPonytailSettingsStore{
142	-	failures: map[int]error{2: errors.New("rollback disk full")},
143	-	}
144	-	refreshCalls := 0
145	-	interactive := NewInteractive(InteractiveConfig{
146	-	SettingsStore: store,
147	-	RefreshPrompt: func() error {
148	-	refreshCalls++
149	-	if refreshCalls == 1 {
150	-	return errors.New("refresh failed")
151	-	}
152	-	return nil
153	-	},
154	-	})
155	-	interactive.rend = nil
156	-
157	-	interactive.applySettingToggle("ponytail_enabled", false)
158	-	if interactive.ponytailEnabled() {
159	-	t.Fatal("in-memory setting reverted despite rollback persistence failure")
160	-	}
161	-	if len(store.values) != 1 \|\| store.values[0] {
162	-	t.Fatalf("durable writes = %v, want only the successful disable", store.values)
163	-	}
164	-	if refreshCalls != 2 {
165	-	t.Fatalf("refresh calls = %d, want failed refresh plus durable-state reconciliation", refreshCalls)
166	-	}
167	-	if !strings.Contains(interactive.statusErr, "rollback disk full") {
168	-	t.Fatalf("status error = %q, want rollback failure", interactive.statusErr)
169	-	}
170	-	}
171	-
172	-	func TestApplyPonytailToggleKeepsStateWhenPersistenceFails(t *testing.T) {
173	-	store := &recordingPonytailSettingsStore{err: errors.New("disk full")}
174	-	refreshCalls := 0
175	-	interactive := NewInteractive(InteractiveConfig{
176	-	SettingsStore: store,
177	-	RefreshPrompt: func() error {
178	-	refreshCalls++
179	-	return nil
180	-	},
181	-	})
182	-	interactive.rend = nil
183	-
184	-	interactive.applySettingToggle("ponytail_enabled", false)
185	-	if !interactive.ponytailEnabled() {
186	-	t.Fatal("Ponytail mode changed despite persistence failure")
187	-	}
188	-	if refreshCalls != 0 {
189	-	t.Fatalf("refresh calls = %d, want none", refreshCalls)
190	-	}
191	-	if !strings.Contains(interactive.statusErr, "disk full") {
192	-	t.Fatalf("status error = %q, want persistence failure", interactive.statusErr)
193	-	}
194	-	}
195	-
196	-	func TestApplyPonytailToggleUpdatesLiveAgentPrompt(t *testing.T) {
197	-	const ponytailBlock = "[ponytail test block]"
198	-	store := &recordingPonytailSettingsStore{}
199	-	ag := core.NewAgent(nil, "model", "base\n\n"+ponytailBlock, nil)
200	-	var interactive *Interactive
201	-	refreshCalls := 0
202	-	interactive = NewInteractive(InteractiveConfig{
203	-	Agent: ag,
204	-	SettingsStore: store,
205	-	RefreshPrompt: func() error {
206	-	refreshCalls++
207	-	system := "base"
208	-	if interactive.ponytailEnabled() {
209	-	system += "\n\n" + ponytailBlock
210	-	}
211	-	ag.SetSystemPrompt(system)
212	-	return nil
213	-	},
214	-	})
215	-	interactive.rend = nil
216	-
217	-	interactive.applySettingToggle("ponytail_enabled", false)
218	-	system, _ := ag.PromptConfig()
219	-	if strings.Contains(system, ponytailBlock) {
220	-	t.Fatalf("live prompt still contains disabled Ponytail block: %q", system)
221	-	}
222	-	interactive.applySettingToggle("ponytail_enabled", true)
223	-	system, _ = ag.PromptConfig()
224	-	if strings.Count(system, ponytailBlock) != 1 {
225	-	t.Fatalf("live prompt block count = %d, want 1: %q", strings.Count(system, ponytailBlock), system)
226	-	}
227	-	if refreshCalls != 2 {
228	-	t.Fatalf("refresh calls = %d, want 2", refreshCalls)
229	-	}
230	-	}

D packages/agent/ponytail.go +0 -15

@@ -1,15 +0,0 @@
1	-	package agent
2	-
3	-	import (
4	-	_ "embed"
5	-	"strings"
6	-	)
7	-
8	-	//go:embed ponytail.md
9	-	var ponytailSystemAddendum string
10	-
11	-	// PonytailSystemAddendum returns the compact coding-guidance block included
12	-	// in resolved system prompts when Ponytail mode is enabled.
13	-	func PonytailSystemAddendum() string {
14	-	return strings.TrimSpace(ponytailSystemAddendum)
15	-	}

@@ -1,15 +0,0 @@
1	-	package agent
2	-
3	-	import (
4	-	_ "embed"
5	-	"strings"
6	-	)
7	-
8	-	//go:embed ponytail.md
9	-	var ponytailSystemAddendum string
10	-
11	-	// PonytailSystemAddendum returns the compact coding-guidance block included
12	-	// in resolved system prompts when Ponytail mode is enabled.
13	-	func PonytailSystemAddendum() string {
14	-	return strings.TrimSpace(ponytailSystemAddendum)
15	-	}

D packages/agent/ponytail.md +0 -14

@@ -1,14 +0,0 @@
1	-	# Ponytail coding mode
2	-
3	-	When coding, debugging, reviewing, or changing software:
4	-
5	-	- Understand the real flow and the owning boundary before editing.
6	-	- Prefer the smallest correct implementation; follow YAGNI.
7	-	- Reuse existing code and patterns before introducing abstractions.
8	-	- Prefer the standard library and native platform features before dependencies.
9	-	- Fix root causes rather than masking symptoms.
10	-	- Keep validation, error handling, security, accessibility, data-loss protection, and runnable checks intact.
11	-	- Avoid speculative features, boilerplate, and unrelated refactors.
12	-	- Before finishing non-trivial work, run the narrowest useful check and report what remains uncertain.
13	-
14	-	This is coding guidance, not a permission or security policy. Apply it to engineering work, not ordinary conversation or creative requests.

@@ -1,14 +0,0 @@
1	-	# Ponytail coding mode
2	-
3	-	When coding, debugging, reviewing, or changing software:
4	-
5	-	- Understand the real flow and the owning boundary before editing.
6	-	- Prefer the smallest correct implementation; follow YAGNI.
7	-	- Reuse existing code and patterns before introducing abstractions.
8	-	- Prefer the standard library and native platform features before dependencies.
9	-	- Fix root causes rather than masking symptoms.
10	-	- Keep validation, error handling, security, accessibility, data-loss protection, and runnable checks intact.
11	-	- Avoid speculative features, boilerplate, and unrelated refactors.
12	-	- Before finishing non-trivial work, run the narrowest useful check and report what remains uncertain.
13	-
14	-	This is coding guidance, not a permission or security policy. Apply it to engineering work, not ordinary conversation or creative requests.

D packages/agent/ponytail_test.go +0 -199

@@ -1,199 +0,0 @@
1	-	package agent
2	-
3	-	import (
4	-	"os"
5	-	"path/filepath"
6	-	"strings"
7	-	"testing"
8	-	)
9	-
10	-	func TestConfigPonytailModeDefaultsToEnabled(t *testing.T) {
11	-	cfg := Config{}
12	-	if !cfg.PonytailModeEnabled() {
13	-	t.Fatal("Ponytail mode should default to enabled")
14	-	}
15	-
16	-	disabled := false
17	-	cfg.PonytailEnabled = &disabled
18	-	if cfg.PonytailModeEnabled() {
19	-	t.Fatal("Ponytail mode should honor an explicit false value")
20	-	}
21	-
22	-	enabled := true
23	-	cfg.PonytailEnabled = &enabled
24	-	if !cfg.PonytailModeEnabled() {
25	-	t.Fatal("Ponytail mode should honor an explicit true value")
26	-	}
27	-	}
28	-
29	-	func TestPonytailSystemAddendumIsCompact(t *testing.T) {
30	-	addendum := PonytailSystemAddendum()
31	-	if addendum == "" {
32	-	t.Fatal("Ponytail system addendum is empty")
33	-	}
34	-	if len(addendum) > 2000 {
35	-	t.Fatalf("Ponytail system addendum is %d bytes, want at most 2000", len(addendum))
36	-	}
37	-	for _, want := range []string{"YAGNI", "standard library", "validation", "ordinary conversation"} {
38	-	if !strings.Contains(addendum, want) {
39	-	t.Fatalf("Ponytail system addendum missing %q:\n%s", want, addendum)
40	-	}
41	-	}
42	-	}
43	-
44	-	func TestResolvePonytailModeDefaultAndExplicitDisable(t *testing.T) {
45	-	cases := []struct {
46	-	name string
47	-	disabled bool
48	-	want bool
49	-	}{
50	-	{name: "default enabled", want: true},
51	-	{name: "explicitly disabled", disabled: true, want: false},
52	-	}
53	-
54	-	for _, tc := range cases {
55	-	t.Run(tc.name, func(t *testing.T) {
56	-	root := t.TempDir()
57	-	t.Setenv("ZUT_HOME", filepath.Join(root, "zut-home"))
58	-	t.Setenv("HOME", filepath.Join(root, "home"))
59	-	if tc.disabled {
60	-	disabled := false
61	-	if err := SaveConfig(Config{PonytailEnabled: &disabled}); err != nil {
62	-	t.Fatal(err)
63	-	}
64	-	}
65	-
66	-	resolved, err := Resolve(Args{
67	-	Provider: "ollama",
68	-	Model: "any-local-model",
69	-	CWD: filepath.Join(root, "project"),
70	-	NoLSP: true,
71	-	NoSkill: true,
72	-	NoTools: true,
73	-	}, false)
74	-	if err != nil {
75	-	t.Fatalf("Resolve failed: %v", err)
76	-	}
77	-	got := strings.Contains(resolved.SystemPrompt, PonytailSystemAddendum())
78	-	if got != tc.want {
79	-	t.Fatalf("Ponytail prompt inclusion = %v, want %v:\n%s", got, tc.want, resolved.SystemPrompt)
80	-	}
81	-	if count := strings.Count(resolved.SystemPrompt, PonytailSystemAddendum()); count != btoi(tc.want) {
82	-	t.Fatalf("Ponytail prompt occurrence count = %d, want %d", count, btoi(tc.want))
83	-	}
84	-	})
85	-	}
86	-	}
87	-
88	-	func btoi(value bool) int {
89	-	if value {
90	-	return 1
91	-	}
92	-	return 0
93	-	}
94	-
95	-	func TestResolvePonytailModeRetainsAddendumWithSystemFile(t *testing.T) {
96	-	root := t.TempDir()
97	-	zutHome := filepath.Join(root, "zut-home")
98	-	t.Setenv("ZUT_HOME", zutHome)
99	-	t.Setenv("HOME", filepath.Join(root, "home"))
100	-	if err := os.MkdirAll(zutHome, 0o755); err != nil {
101	-	t.Fatal(err)
102	-	}
103	-	if err := os.WriteFile(filepath.Join(zutHome, "SYSTEM.md"), []byte("custom system identity"), 0o600); err != nil {
104	-	t.Fatal(err)
105	-	}
106	-
107	-	resolved, err := Resolve(Args{
108	-	Provider: "ollama",
109	-	Model: "any-local-model",
110	-	CWD: filepath.Join(root, "project"),
111	-	NoLSP: true,
112	-	NoSkill: true,
113	-	NoTools: true,
114	-	}, false)
115	-	if err != nil {
116	-	t.Fatalf("Resolve failed: %v", err)
117	-	}
118	-	if !strings.Contains(resolved.SystemPrompt, "custom system identity") {
119	-	t.Fatalf("SYSTEM.md content missing:\n%s", resolved.SystemPrompt)
120	-	}
121	-	if !strings.Contains(resolved.SystemPrompt, PonytailSystemAddendum()) {
122	-	t.Fatalf("SYSTEM.md replacement lost Ponytail addendum:\n%s", resolved.SystemPrompt)
123	-	}
124	-	}
125	-
126	-	func TestResolvePonytailModePrecedesExplicitAppendPrompt(t *testing.T) {
127	-	root := t.TempDir()
128	-	t.Setenv("ZUT_HOME", filepath.Join(root, "zut-home"))
129	-	t.Setenv("HOME", filepath.Join(root, "home"))
130	-
131	-	const explicit = "explicit invocation guidance"
132	-	resolved, err := Resolve(Args{
133	-	Provider: "ollama",
134	-	Model: "any-local-model",
135	-	CWD: filepath.Join(root, "project"),
136	-	NoLSP: true,
137	-	NoSkill: true,
138	-	NoTools: true,
139	-	AppendSystemPrompt: []string{explicit},
140	-	}, false)
141	-	if err != nil {
142	-	t.Fatalf("Resolve failed: %v", err)
143	-	}
144	-	ponytailIndex := strings.Index(resolved.SystemPrompt, PonytailSystemAddendum())
145	-	explicitIndex := strings.Index(resolved.SystemPrompt, explicit)
146	-	if ponytailIndex < 0 \|\| explicitIndex < 0 \|\| ponytailIndex > explicitIndex {
147	-	t.Fatalf("prompt ordering = Ponytail %d, explicit %d:\n%s", ponytailIndex, explicitIndex, resolved.SystemPrompt)
148	-	}
149	-	}
150	-
151	-	func TestResolvePonytailModeAppliesAcrossHeadlessModesAndCustomIdentity(t *testing.T) {
152	-	root := t.TempDir()
153	-	t.Setenv("ZUT_HOME", filepath.Join(root, "zut-home"))
154	-	t.Setenv("HOME", filepath.Join(root, "home"))
155	-
156	-	modes := []Mode{"", ModePrint, ModeStream, ModeJSON, ModeRPC}
157	-	for _, mode := range modes {
158	-	resolved, err := Resolve(Args{
159	-	Mode: mode,
160	-	Provider: "ollama",
161	-	Model: "any-local-model",
162	-	CWD: filepath.Join(root, "project"),
163	-	NoLSP: true,
164	-	NoSkill: true,
165	-	NoTools: true,
166	-	SystemPrompt: "custom identity",
167	-	}, false)
168	-	if err != nil {
169	-	t.Fatalf("Resolve(%q) failed: %v", mode, err)
170	-	}
171	-	if !strings.Contains(resolved.SystemPrompt, "custom identity") {
172	-	t.Fatalf("Resolve(%q) lost custom identity:\n%s", mode, resolved.SystemPrompt)
173	-	}
174	-	if !strings.Contains(resolved.SystemPrompt, PonytailSystemAddendum()) {
175	-	t.Fatalf("Resolve(%q) omitted Ponytail guidance:\n%s", mode, resolved.SystemPrompt)
176	-	}
177	-	}
178	-	}
179	-
180	-	func TestConfigSettingsStorePersistsPonytailWithoutChangingKnownFields(t *testing.T) {
181	-	t.Setenv("ZUT_HOME", t.TempDir())
182	-	if err := SaveConfig(Config{Theme: "dark", Provider: "ollama", Model: "local"}); err != nil {
183	-	t.Fatal(err)
184	-	}
185	-
186	-	if err := (configSettingsStore{}).SetPonytailEnabled(false); err != nil {
187	-	t.Fatal(err)
188	-	}
189	-	cfg, err := LoadConfig()
190	-	if err != nil {
191	-	t.Fatal(err)
192	-	}
193	-	if cfg.PonytailEnabled == nil \|\| *cfg.PonytailEnabled {
194	-	t.Fatal("Ponytail mode was not persisted as disabled")
195	-	}
196	-	if cfg.Theme != "dark" \|\| cfg.Provider != "ollama" \|\| cfg.Model != "local" {
197	-	t.Fatalf("known config fields changed: %#v", cfg)
198	-	}
199	-	}

@@ -1,199 +0,0 @@
1	-	package agent
2	-
3	-	import (
4	-	"os"
5	-	"path/filepath"
6	-	"strings"
7	-	"testing"
8	-	)
9	-
10	-	func TestConfigPonytailModeDefaultsToEnabled(t *testing.T) {
11	-	cfg := Config{}
12	-	if !cfg.PonytailModeEnabled() {
13	-	t.Fatal("Ponytail mode should default to enabled")
14	-	}
15	-
16	-	disabled := false
17	-	cfg.PonytailEnabled = &disabled
18	-	if cfg.PonytailModeEnabled() {
19	-	t.Fatal("Ponytail mode should honor an explicit false value")
20	-	}
21	-
22	-	enabled := true
23	-	cfg.PonytailEnabled = &enabled
24	-	if !cfg.PonytailModeEnabled() {
25	-	t.Fatal("Ponytail mode should honor an explicit true value")
26	-	}
27	-	}
28	-
29	-	func TestPonytailSystemAddendumIsCompact(t *testing.T) {
30	-	addendum := PonytailSystemAddendum()
31	-	if addendum == "" {
32	-	t.Fatal("Ponytail system addendum is empty")
33	-	}
34	-	if len(addendum) > 2000 {
35	-	t.Fatalf("Ponytail system addendum is %d bytes, want at most 2000", len(addendum))
36	-	}
37	-	for _, want := range []string{"YAGNI", "standard library", "validation", "ordinary conversation"} {
38	-	if !strings.Contains(addendum, want) {
39	-	t.Fatalf("Ponytail system addendum missing %q:\n%s", want, addendum)
40	-	}
41	-	}
42	-	}
43	-
44	-	func TestResolvePonytailModeDefaultAndExplicitDisable(t *testing.T) {
45	-	cases := []struct {
46	-	name string
47	-	disabled bool
48	-	want bool
49	-	}{
50	-	{name: "default enabled", want: true},
51	-	{name: "explicitly disabled", disabled: true, want: false},
52	-	}
53	-
54	-	for _, tc := range cases {
55	-	t.Run(tc.name, func(t *testing.T) {
56	-	root := t.TempDir()
57	-	t.Setenv("ZUT_HOME", filepath.Join(root, "zut-home"))
58	-	t.Setenv("HOME", filepath.Join(root, "home"))
59	-	if tc.disabled {
60	-	disabled := false
61	-	if err := SaveConfig(Config{PonytailEnabled: &disabled}); err != nil {
62	-	t.Fatal(err)
63	-	}
64	-	}
65	-
66	-	resolved, err := Resolve(Args{
67	-	Provider: "ollama",
68	-	Model: "any-local-model",
69	-	CWD: filepath.Join(root, "project"),
70	-	NoLSP: true,
71	-	NoSkill: true,
72	-	NoTools: true,
73	-	}, false)
74	-	if err != nil {
75	-	t.Fatalf("Resolve failed: %v", err)
76	-	}
77	-	got := strings.Contains(resolved.SystemPrompt, PonytailSystemAddendum())
78	-	if got != tc.want {
79	-	t.Fatalf("Ponytail prompt inclusion = %v, want %v:\n%s", got, tc.want, resolved.SystemPrompt)
80	-	}
81	-	if count := strings.Count(resolved.SystemPrompt, PonytailSystemAddendum()); count != btoi(tc.want) {
82	-	t.Fatalf("Ponytail prompt occurrence count = %d, want %d", count, btoi(tc.want))
83	-	}
84	-	})
85	-	}
86	-	}
87	-
88	-	func btoi(value bool) int {
89	-	if value {
90	-	return 1
91	-	}
92	-	return 0
93	-	}
94	-
95	-	func TestResolvePonytailModeRetainsAddendumWithSystemFile(t *testing.T) {
96	-	root := t.TempDir()
97	-	zutHome := filepath.Join(root, "zut-home")
98	-	t.Setenv("ZUT_HOME", zutHome)
99	-	t.Setenv("HOME", filepath.Join(root, "home"))
100	-	if err := os.MkdirAll(zutHome, 0o755); err != nil {
101	-	t.Fatal(err)
102	-	}
103	-	if err := os.WriteFile(filepath.Join(zutHome, "SYSTEM.md"), []byte("custom system identity"), 0o600); err != nil {
104	-	t.Fatal(err)
105	-	}
106	-
107	-	resolved, err := Resolve(Args{
108	-	Provider: "ollama",
109	-	Model: "any-local-model",
110	-	CWD: filepath.Join(root, "project"),
111	-	NoLSP: true,
112	-	NoSkill: true,
113	-	NoTools: true,
114	-	}, false)
115	-	if err != nil {
116	-	t.Fatalf("Resolve failed: %v", err)
117	-	}
118	-	if !strings.Contains(resolved.SystemPrompt, "custom system identity") {
119	-	t.Fatalf("SYSTEM.md content missing:\n%s", resolved.SystemPrompt)
120	-	}
121	-	if !strings.Contains(resolved.SystemPrompt, PonytailSystemAddendum()) {
122	-	t.Fatalf("SYSTEM.md replacement lost Ponytail addendum:\n%s", resolved.SystemPrompt)
123	-	}
124	-	}
125	-
126	-	func TestResolvePonytailModePrecedesExplicitAppendPrompt(t *testing.T) {
127	-	root := t.TempDir()
128	-	t.Setenv("ZUT_HOME", filepath.Join(root, "zut-home"))
129	-	t.Setenv("HOME", filepath.Join(root, "home"))
130	-
131	-	const explicit = "explicit invocation guidance"
132	-	resolved, err := Resolve(Args{
133	-	Provider: "ollama",
134	-	Model: "any-local-model",
135	-	CWD: filepath.Join(root, "project"),
136	-	NoLSP: true,
137	-	NoSkill: true,
138	-	NoTools: true,
139	-	AppendSystemPrompt: []string{explicit},
140	-	}, false)
141	-	if err != nil {
142	-	t.Fatalf("Resolve failed: %v", err)
143	-	}
144	-	ponytailIndex := strings.Index(resolved.SystemPrompt, PonytailSystemAddendum())
145	-	explicitIndex := strings.Index(resolved.SystemPrompt, explicit)
146	-	if ponytailIndex < 0 \|\| explicitIndex < 0 \|\| ponytailIndex > explicitIndex {
147	-	t.Fatalf("prompt ordering = Ponytail %d, explicit %d:\n%s", ponytailIndex, explicitIndex, resolved.SystemPrompt)
148	-	}
149	-	}
150	-
151	-	func TestResolvePonytailModeAppliesAcrossHeadlessModesAndCustomIdentity(t *testing.T) {
152	-	root := t.TempDir()
153	-	t.Setenv("ZUT_HOME", filepath.Join(root, "zut-home"))
154	-	t.Setenv("HOME", filepath.Join(root, "home"))
155	-
156	-	modes := []Mode{"", ModePrint, ModeStream, ModeJSON, ModeRPC}
157	-	for _, mode := range modes {
158	-	resolved, err := Resolve(Args{
159	-	Mode: mode,
160	-	Provider: "ollama",
161	-	Model: "any-local-model",
162	-	CWD: filepath.Join(root, "project"),
163	-	NoLSP: true,
164	-	NoSkill: true,
165	-	NoTools: true,
166	-	SystemPrompt: "custom identity",
167	-	}, false)
168	-	if err != nil {
169	-	t.Fatalf("Resolve(%q) failed: %v", mode, err)
170	-	}
171	-	if !strings.Contains(resolved.SystemPrompt, "custom identity") {
172	-	t.Fatalf("Resolve(%q) lost custom identity:\n%s", mode, resolved.SystemPrompt)
173	-	}
174	-	if !strings.Contains(resolved.SystemPrompt, PonytailSystemAddendum()) {
175	-	t.Fatalf("Resolve(%q) omitted Ponytail guidance:\n%s", mode, resolved.SystemPrompt)
176	-	}
177	-	}
178	-	}
179	-
180	-	func TestConfigSettingsStorePersistsPonytailWithoutChangingKnownFields(t *testing.T) {
181	-	t.Setenv("ZUT_HOME", t.TempDir())
182	-	if err := SaveConfig(Config{Theme: "dark", Provider: "ollama", Model: "local"}); err != nil {
183	-	t.Fatal(err)
184	-	}
185	-
186	-	if err := (configSettingsStore{}).SetPonytailEnabled(false); err != nil {
187	-	t.Fatal(err)
188	-	}
189	-	cfg, err := LoadConfig()
190	-	if err != nil {
191	-	t.Fatal(err)
192	-	}
193	-	if cfg.PonytailEnabled == nil \|\| *cfg.PonytailEnabled {
194	-	t.Fatal("Ponytail mode was not persisted as disabled")
195	-	}
196	-	if cfg.Theme != "dark" \|\| cfg.Provider != "ollama" \|\| cfg.Model != "local" {
197	-	t.Fatalf("known config fields changed: %#v", cfg)
198	-	}
199	-	}

M packages/agent/settings_store.go +0 -12


@@ -106,18 +106,6 @@ func (configSettingsStore) SetAutoSubagents(enabled bool) error {
106	106		return SaveConfig(cfg)
107	107		}
108	108
109		-	func (configSettingsStore) SetPonytailEnabled(enabled bool) error {
110		-	cfg, err := LoadConfig()
111		-	if err != nil {
112		-	return fmt.Errorf("load config for Ponytail setting: %w", err)
113		-	}
114		-	cfg.PonytailEnabled = &enabled
115		-	if err := SaveConfig(cfg); err != nil {
116		-	return fmt.Errorf("save Ponytail setting: %w", err)
117		-	}
118		-	return nil
119		-	}
120		-
121	109		func (configSettingsStore) SetWebSearchEnabled(enabled bool) error {
122	110		cfg, err := LoadConfig()
123	111		if err != nil {


@@ -106,18 +106,6 @@ func (configSettingsStore) SetAutoSubagents(enabled bool) error {
106		return SaveConfig(cfg)	106	return SaveConfig(cfg)
107		}	107	}
108			108
109	-	func (configSettingsStore) SetPonytailEnabled(enabled bool) error {
110	-	cfg, err := LoadConfig()
111	-	if err != nil {
112	-	return fmt.Errorf("load config for Ponytail setting: %w", err)
113	-	}
114	-	cfg.PonytailEnabled = &enabled
115	-	if err := SaveConfig(cfg); err != nil {
116	-	return fmt.Errorf("save Ponytail setting: %w", err)
117	-	}
118	-	return nil
119	-	}
120	-
121		func (configSettingsStore) SetWebSearchEnabled(enabled bool) error {	109	func (configSettingsStore) SetWebSearchEnabled(enabled bool) error {
122		cfg, err := LoadConfig()	110	cfg, err := LoadConfig()
123		if err != nil {	111	if err != nil {

M packages/agent/systemprompt.go +1 -1


@@ -109,7 +109,7 @@ const toolSelectionGuidance = `Prefer dedicated tools over bash: glob for file d
109	109
110	110		const compactedSummaryHandoffInstruction = `When you see a "## Context Summary (compacted)" message, treat it as a handoff from earlier work. Keep its active constraints and preferences in force, continue the most recent unresolved user request without waiting for the user to type "continue", and follow a newer user request when it supersedes the summary.`
111	111
112		-	const taskExecutionGuidance = `Treat requests such as "can you" as instructions to do the requested work within the user's intended scope; do not stop at a plan or an offer to continue. Proceed with authorized, non-destructive work and reasonable assumptions until the task is complete or genuinely blocked. Follow each progress statement with the action it describes: pair every intermediate update with tool calls or a concrete next step in the same turn. Final answers contain the result of the work or a concrete blocker, never a bare promise to continue. Respect requests for explanation, planning, or review without making unsolicited changes. Ask only when a necessary decision blocks safe, correct progress; complete independent authorized work first. Preserve tool permissions, required confirmations, and explicit approval requirements; do not infer permission for destructive or irreversible actions.`
	112	+	const taskExecutionGuidance = `Treat requests such as "can you" as instructions to do the requested work within the user's intended scope; do not stop at a plan or an offer to continue. Proceed with authorized, non-destructive work and reasonable assumptions until the task is complete or genuinely blocked. Follow each progress statement with the action it describes: pair every intermediate update with tool calls or a concrete next step in the same turn. Final answers contain the result of the work or a concrete blocker, never a bare promise to continue. Respect requests for explanation, planning, or review without making unsolicited changes. Ask only when a necessary decision blocks safe, correct progress; complete independent authorized work first. Verify changes with checks appropriate to their scope. Report which checks ran and what remains unverified. Preserve tool permissions, required confirmations, and explicit approval requirements; do not infer permission for destructive or irreversible actions.`
113	113
114	114		const planGuidance = `For multi-phase or dependency-ordered work, and whenever the user asks for a plan, maintain the plan checklist instead of tracking steps in prose. Skip it for single-step or informational requests. Keep at most one step in_progress and update the plan at phase boundaries, not after every command. Group related steps instead of appending endlessly: a plan is phases you can verify, not a task log. Never mark a step completed without doing the work, and do not batch-complete steps after the fact. Prefer add, update, and remove over re-sending the whole list with set; use set only to replace the checklist and clear to remove it. After a plan call, do not restate the plan in prose because the interface already shows it. After compaction, or whenever step identity is unclear, call plan with action:"show" instead of guessing an index. A plan carried over from earlier work stays visible until it is cleared or replaced, so clear it once the work it describes is no longer in flight.`
115	115


@@ -109,7 +109,7 @@ const toolSelectionGuidance = `Prefer dedicated tools over bash: glob for file d
109			109
110		const compactedSummaryHandoffInstruction = `When you see a "## Context Summary (compacted)" message, treat it as a handoff from earlier work. Keep its active constraints and preferences in force, continue the most recent unresolved user request without waiting for the user to type "continue", and follow a newer user request when it supersedes the summary.`	110		const compactedSummaryHandoffInstruction = `When you see a "## Context Summary (compacted)" message, treat it as a handoff from earlier work. Keep its active constraints and preferences in force, continue the most recent unresolved user request without waiting for the user to type "continue", and follow a newer user request when it supersedes the summary.`
111			111
112	-	const taskExecutionGuidance = `Treat requests such as "can you" as instructions to do the requested work within the user's intended scope; do not stop at a plan or an offer to continue. Proceed with authorized, non-destructive work and reasonable assumptions until the task is complete or genuinely blocked. Follow each progress statement with the action it describes: pair every intermediate update with tool calls or a concrete next step in the same turn. Final answers contain the result of the work or a concrete blocker, never a bare promise to continue. Respect requests for explanation, planning, or review without making unsolicited changes. Ask only when a necessary decision blocks safe, correct progress; complete independent authorized work first. Preserve tool permissions, required confirmations, and explicit approval requirements; do not infer permission for destructive or irreversible actions.`	112	+	const taskExecutionGuidance = `Treat requests such as "can you" as instructions to do the requested work within the user's intended scope; do not stop at a plan or an offer to continue. Proceed with authorized, non-destructive work and reasonable assumptions until the task is complete or genuinely blocked. Follow each progress statement with the action it describes: pair every intermediate update with tool calls or a concrete next step in the same turn. Final answers contain the result of the work or a concrete blocker, never a bare promise to continue. Respect requests for explanation, planning, or review without making unsolicited changes. Ask only when a necessary decision blocks safe, correct progress; complete independent authorized work first. Verify changes with checks appropriate to their scope. Report which checks ran and what remains unverified. Preserve tool permissions, required confirmations, and explicit approval requirements; do not infer permission for destructive or irreversible actions.`
113			113
114		const planGuidance = `For multi-phase or dependency-ordered work, and whenever the user asks for a plan, maintain the plan checklist instead of tracking steps in prose. Skip it for single-step or informational requests. Keep at most one step in_progress and update the plan at phase boundaries, not after every command. Group related steps instead of appending endlessly: a plan is phases you can verify, not a task log. Never mark a step completed without doing the work, and do not batch-complete steps after the fact. Prefer add, update, and remove over re-sending the whole list with set; use set only to replace the checklist and clear to remove it. After a plan call, do not restate the plan in prose because the interface already shows it. After compaction, or whenever step identity is unclear, call plan with action:"show" instead of guessing an index. A plan carried over from earlier work stays visible until it is cleared or replaced, so clear it once the work it describes is no longer in flight.`	114		const planGuidance = `For multi-phase or dependency-ordered work, and whenever the user asks for a plan, maintain the plan checklist instead of tracking steps in prose. Skip it for single-step or informational requests. Keep at most one step in_progress and update the plan at phase boundaries, not after every command. Group related steps instead of appending endlessly: a plan is phases you can verify, not a task log. Never mark a step completed without doing the work, and do not batch-complete steps after the fact. Prefer add, update, and remove over re-sending the whole list with set; use set only to replace the checklist and clear to remove it. After a plan call, do not restate the plan in prose because the interface already shows it. After compaction, or whenever step identity is unclear, call plan with action:"show" instead of guessing an index. A plan carried over from earlier work stays visible until it is cleared or replaced, so clear it once the work it describes is no longer in flight.`
115			115

M packages/agent/systemprompt_test.go +2 -0


@@ -113,6 +113,8 @@ func TestBuildSystemPromptAlwaysIncludesTaskAndSkillGuidance(t *testing.T) {
113	113		"Final answers contain the result of the work or a concrete blocker",
114	114		"Respect requests for explanation, planning, or review without making unsolicited changes",
115	115		"Ask only when a necessary decision blocks safe, correct progress",
	116	+	"Verify changes with checks appropriate to their scope.",
	117	+	"Report which checks ran and what remains unverified.",
116	118		"Preserve tool permissions, required confirmations, and explicit approval requirements",
117	119		"Explicit user instructions take precedence over skill guidelines",
118	120		"Skills do not grant permissions or override system or developer constraints",


@@ -113,6 +113,8 @@ func TestBuildSystemPromptAlwaysIncludesTaskAndSkillGuidance(t *testing.T) {
113	"Final answers contain the result of the work or a concrete blocker",	113		"Final answers contain the result of the work or a concrete blocker",
114	"Respect requests for explanation, planning, or review without making unsolicited changes",	114		"Respect requests for explanation, planning, or review without making unsolicited changes",
115	"Ask only when a necessary decision blocks safe, correct progress",	115		"Ask only when a necessary decision blocks safe, correct progress",
		116	+	"Verify changes with checks appropriate to their scope.",
		117	+	"Report which checks ran and what remains unverified.",
116	"Preserve tool permissions, required confirmations, and explicit approval requirements",	118		"Preserve tool permissions, required confirmations, and explicit approval requirements",
117	"Explicit user instructions take precedence over skill guidelines",	119		"Explicit user instructions take precedence over skill guidelines",
118	"Skills do not grant permissions or override system or developer constraints",	120		"Skills do not grant permissions or override system or developer constraints",