fix: make ponytail mode changes visible to the user on Claude Code (#648)

.openclaw/skills/ponytail-help/SKILL.md

@@ -42,7 +42,13 @@ Say "stop ponytail" or "normal mode". Resume anytime with `/ponytail`.
 
 ## Configure Default Mode
 
-Default mode = `full`, auto-active every session. Change it:
+Default mode = `full`, auto-active every session. `/ponytail off` lasts only for
+the current session — the next one starts at the default again. Change it:
+
+**Command**:
+```
+/ponytail default lite
+```
 
 **Environment variable** (highest priority):
 ```bash

.openclaw/skills/ponytail/SKILL.md

@@ -15,7 +15,8 @@ code is the code never written.
 
 ACTIVE EVERY RESPONSE. No drift back to over-building. Still active if
 unsure. Off only: "stop ponytail" / "normal mode". Default: **full**.
-Switch: `/ponytail lite|full|ultra`.
+Switch: `/ponytail lite|full|ultra|off`. Bare `/ponytail` reports the level.
+`/ponytail default <mode>` persists it to new sessions.
 
 ## The ladder
 

README.es.md

@@ -234,7 +234,8 @@ Qué archivos corresponden a qué agente: [Portabilidad de agentes](docs/agent-p
 
 | Comando | Qué hace |
 |---------|----------|
-| `/ponytail [lite \| full \| ultra \| off]` | Cambia la intensidad, o apágalo. Sin argumento, reporta el nivel actual. |
+| `/ponytail [lite \| full \| ultra \| off]` | Cambia la intensidad, o apágalo. Sin argumento, reporta el nivel actual. Dura solo la sesión: una sesión nueva arranca en el valor por defecto. |
+| `/ponytail default [lite \| full \| ultra \| off]` | Guarda el valor por defecto, para que las sesiones nuevas arranquen ahí. |
 | `/ponytail-review` | Revisa el diff actual en busca de sobre-ingeniería y devuelve una lista de qué eliminar. |
 | `/ponytail-audit` | Audita el repo completo en busca de sobre-ingeniería, no solo el diff. |
 | `/ponytail-debt` | Recolecta los atajos marcados con `ponytail:` que dejaste pendientes en un registro, para que "después" no se convierta en "nunca". |

README.ko.md

@@ -253,7 +253,8 @@ Codex 확장을 쓰는 VS Code는 이 저장소가 함께 싣는 `AGENTS.md`를
 
 | 명령 | 하는 일 |
 |---------|--------------|
-| `/ponytail [lite \| full \| ultra \| off]` | 강도를 정하거나, 끈다. 인수가 없으면 지금 레벨을 알려 준다. |
+| `/ponytail [lite \| full \| ultra \| off]` | 강도를 정하거나, 끈다. 인수가 없으면 지금 레벨을 알려 준다. 세션 한정이라, 새 세션은 기본값에서 시작한다. |
+| `/ponytail default [lite \| full \| ultra \| off]` | 기본값을 저장해서, 새 세션이 그 레벨에서 시작하게 한다. |
 | `/ponytail-review` | 지금 diff를 과잉 구현 관점에서 훑고, 삭제 목록을 돌려준다. |
 | `/ponytail-audit` | diff만이 아니라 저장소 전체를 과잉 구현 관점에서 감사한다. |
 | `/ponytail-debt` | 미뤄 둔 `ponytail:` 간소화들을 장부로 모아, "나중에"가 "영영"이 되지 않게 한다. |

README.md

@@ -286,7 +286,8 @@ These remove the plugin's own files. They leave behind a small amount of state p
 
 | Command | What it does |
 |---------|--------------|
-| `/ponytail [lite \| full \| ultra \| off]` | Set the intensity, or turn it off. No argument reports the current level. |
+| `/ponytail [lite \| full \| ultra \| off]` | Set the intensity, or turn it off. No argument reports the current level. Session-scoped: a new session starts at the default. |
+| `/ponytail default [lite \| full \| ultra \| off]` | Persist the default, so new sessions start there. |
 | `/ponytail-review` | Review the current diff for over-engineering, hands back a delete-list. |
 | `/ponytail-audit` | Audit the whole repo for over-engineering, not just the diff. |
 | `/ponytail-debt` | Harvest the `ponytail:` shortcuts you've deferred into a ledger, so "later" doesn't become "never". |

hooks/ponytail-instructions.js

@@ -45,7 +45,8 @@ function getFallbackInstructions(mode) {
     'You are a lazy senior developer. Lazy means efficient, not careless. The best code is the code never written.\n\n' +
     '## Persistence\n\n' +
     'ACTIVE EVERY RESPONSE. No drift back to over-building. Still active if unsure. Off only: "stop ponytail" / "normal mode".\n\n' +
-    'Current level: **' + mode + '**. Switch: `/ponytail lite|full|ultra`.\n\n' +
+    'Current level: **' + mode + '**. Switch: `/ponytail lite|full|ultra|off`. ' +
+    'Bare `/ponytail` reports the level; `/ponytail default <mode>` persists it.\n\n' +
     '## The ladder\n\n' +
     'Before any code, stop at the first rung that holds (the ladder runs after you understand the problem, not instead of it — read the code it touches and trace the real flow first):\n' +
     '1. Does this need to be built at all? (YAGNI)\n' +

hooks/ponytail-mode-tracker.js

@@ -9,6 +9,24 @@ const { getPonytailInstructions } = require('./ponytail-instructions');
 let input = '';
 let done = false;
 
+// Name the default only when it differs from the live level, so the common case
+// stays a three-word status line and the surprising case explains itself.
+function describeDefault(mode) {
+  const fallback = getDefaultMode();
+  return fallback === mode ? '' : ' (default: ' + fallback + ')';
+}
+
+// Off is session-scoped: clearMode() only drops the flag, and the next
+// SessionStart re-activates at the configured default. Say so here, or the
+// user reads tomorrow's active session as ponytail ignoring them (#648).
+function offMessage() {
+  const fallback = getDefaultMode();
+  return fallback === 'off'
+    ? 'Ponytail off.'
+    : 'Ponytail off for this session — default is ' + fallback +
+      ', so a new session starts there. `/ponytail default off` to persist.';
+}
+
 function finish() {
   if (done) return;
   done = true;
@@ -39,7 +57,12 @@ function finish() {
           const dmode = parts[2];
           if (dmode === 'off' || dmode === 'lite' || dmode === 'full' || dmode === 'ultra') {
             writeDefaultMode(dmode);
-            writeHookOutput('UserPromptSubmit', dmode, 'PONYTAIL DEFAULT SET — new sessions start in ' + dmode + '.');
+            writeHookOutput(
+              'UserPromptSubmit',
+              dmode,
+              'PONYTAIL DEFAULT SET — new sessions start in ' + dmode + '.',
+              'Ponytail default set to ' + dmode + ' — new sessions start there.',
+            );
           }
           return; // don't fall through to the session-mode switch
         }
@@ -60,6 +83,7 @@ function finish() {
           'UserPromptSubmit',
           mode,
           'PONYTAIL MODE ACTIVE — level: ' + mode,
+          'Ponytail: ' + mode + describeDefault(mode),
         );
       } else if (mode && mode !== 'off') {
         setMode(mode);
@@ -72,20 +96,21 @@ function finish() {
             'UserPromptSubmit',
             mode,
             'PONYTAIL MODE CHANGED — level: ' + mode,
+            'Ponytail: ' + mode + describeDefault(mode),
           );
         }
       } else if (mode === 'off') {
         clearMode();
         deactivated = true;
-        writeHookOutput('UserPromptSubmit', 'off', 'PONYTAIL MODE OFF');
+        writeHookOutput('UserPromptSubmit', 'off', 'PONYTAIL MODE OFF', offMessage());
       }
     }
 
     // Detect deactivation
     if (!modeSwitched && !deactivated && isDeactivationCommand(prompt)) {
       clearMode();
       deactivated = true;
-      writeHookOutput('UserPromptSubmit', 'off', 'PONYTAIL MODE OFF');
+      writeHookOutput('UserPromptSubmit', 'off', 'PONYTAIL MODE OFF', offMessage());
     }
 
     // Qoder has no SessionStart event, so UserPromptSubmit does double duty:

hooks/ponytail-runtime.js

@@ -33,7 +33,14 @@ function readMode() {
   }
 }
 
-function writeHookOutput(event, mode, context = '') {
+// `userMessage`, when set, is text the user asked for and must actually see.
+// Every other write is context for the model. The distinction matters because on
+// native Claude a UserPromptSubmit hook's stdout becomes additionalContext —
+// injected into the assistant's conversation, never shown — so a mode report
+// written that way reaches everyone except the person who typed the command
+// (#648). Codex already has a systemMessage channel below; this gives the native
+// path the same, without making every hook write user-visible.
+function writeHookOutput(event, mode, context = '', userMessage = '') {
   if (isCopilot) {
     // Copilot reads additionalContext on SessionStart; ignores output elsewhere.
     process.stdout.write(JSON.stringify(
@@ -71,6 +78,14 @@ function writeHookOutput(event, mode, context = '') {
       { hookSpecificOutput: { hookEventName: event, additionalContext: context } }));
     return;
   }
+  if (userMessage) {
+    const output = { systemMessage: userMessage };
+    if (context) {
+      output.hookSpecificOutput = { hookEventName: event, additionalContext: context };
+    }
+    process.stdout.write(JSON.stringify(output));
+    return;
+  }
   process.stdout.write(context);
 }
 

skills/ponytail-help/SKILL.md

@@ -43,7 +43,13 @@ Say "stop ponytail" or "normal mode". Resume anytime with `/ponytail`.
 
 ## Configure Default Mode
 
-Default mode = `full`, auto-active every session. Change it:
+Default mode = `full`, auto-active every session. `/ponytail off` lasts only for
+the current session — the next one starts at the default again. Change it:
+
+**Command**:
+```
+/ponytail default lite
+```
 
 **Environment variable** (highest priority):
 ```bash

skills/ponytail/SKILL.md

@@ -13,7 +13,7 @@ description: >
   over-engineering, bloat, boilerplate, or unnecessary dependencies. Do NOT
   use for non-coding requests (general knowledge, prose, translation,
   summaries, recipes).
-argument-hint: "[lite|full|ultra]"
+argument-hint: "[lite|full|ultra|off]"
 license: MIT
 ---
 
@@ -27,7 +27,8 @@ code is the code never written.
 
 ACTIVE EVERY RESPONSE. No drift back to over-building. Still active if
 unsure. Off only: "stop ponytail" / "normal mode". Default: **full**.
-Switch: `/ponytail lite|full|ultra`.
+Switch: `/ponytail lite|full|ultra|off`. Bare `/ponytail` reports the level.
+`/ponytail default <mode>` persists it to new sessions.
 
 ## The ladder
 

tests/hooks.test.js

@@ -459,4 +459,49 @@ try {
   if (prevEnvModeRev === undefined) delete process.env.PONYTAIL_DEFAULT_MODE; else process.env.PONYTAIL_DEFAULT_MODE = prevEnvModeRev;
 }
 
+// #648: on native Claude a UserPromptSubmit hook's raw stdout becomes
+// additionalContext (model-only), so anything the user asked to see must ride a
+// systemMessage. Report, switch and off all qualify; the ruleset injection that
+// nobody asked to read must stay raw stdout.
+const msgHome = path.join(temp, 'user-message-home');
+const msgEnv = {
+  HOME: msgHome,
+  USERPROFILE: msgHome,
+  XDG_CONFIG_HOME: path.join(msgHome, '.config'),
+};
+
+function systemMessage(prompt) {
+  const r = run('ponytail-mode-tracker.js', msgEnv, JSON.stringify({ prompt }));
+  assert.equal(r.status, 0, r.stderr);
+  if (!r.stdout.startsWith('{')) return null;
+  return JSON.parse(r.stdout).systemMessage || null;
+}
+
+// Switching says which level took effect.
+assert.match(systemMessage('/ponytail ultra') || '', /ultra/, 'a mode switch must be visible to the user');
+
+// Bare /ponytail reports the live level, and does not change it.
+assert.match(systemMessage('/ponytail') || '', /ultra/, 'the level report must be visible to the user');
+assert.equal(fs.readFileSync(path.join(msgHome, '.claude', '.ponytail-active'), 'utf8'), 'ultra',
+  'the report must not change the mode');
+
+// Off names the default it will revert to, since clearMode() is session-scoped.
+const offMsg = systemMessage('/ponytail off') || '';
+assert.match(offMsg, /off/i, 'deactivation must be visible to the user');
+assert.match(offMsg, /default is full/, 'off must name the default that revives ponytail next session');
+
+// Same for the natural-language form.
+assert.match(systemMessage('stop ponytail') || '', /off/i, '"stop ponytail" must be visible to the user');
+
+// A default change is user-visible too, and distinguishable from a switch.
+assert.match(systemMessage('/ponytail default lite') || '', /default set to lite/i,
+  'a default change must be visible to the user');
+
+// The ruleset injection is not a user message: it stays raw stdout so Claude
+// Code keeps treating it as additionalContext.
+const activate = run('ponytail-activate.js', msgEnv);
+assert.equal(activate.status, 0, activate.stderr);
+assert.ok(!activate.stdout.trimStart().startsWith('{'),
+  'SessionStart ruleset must stay raw stdout, not a systemMessage envelope');
+
 console.log('hook compatibility checks passed');