fix(prompts): prefer task_id in continuation guidance

2026-04-18 10:41:31 +08:00
parent 1a8f60b89e
commit 64a2911880
8 changed files with 20 additions and 20 deletions
@@ -139,10 +139,10 @@ Count remaining **top-level task** checkboxes. Ignore nested verification/eviden
 [ ] Boulder: Read plan file, confirmed current progress
 \`\`\`

-**If verification fails**: Resume the SAME session with the ACTUAL error output:
+**If verification fails**: Resume the SAME task with the ACTUAL error output:
 \`\`\`typescript
 task(
-  session_id="ses_xyz789",
+  task_id="ses_xyz789",
  load_skills=[...],
  prompt="Verification failed: {actual error}. Fix."
 )
@@ -156,7 +156,7 @@ Every \`task()\` output includes a task_id. STORE IT.

 If task fails:
 1. Identify what went wrong
-2. **Resume the SAME session** - subagent has full context already:
+2. **Resume the SAME task** - subagent has full context already:
    \`\`\`typescript
    task(
      task_id="ses_xyz789",  // Task ID from failed task
@@ -187,7 +187,7 @@ Final-wave reviewers can finish in parallel before you update the plan file, so

 1. Execute all Final Wave tasks in parallel
 2. If ANY verdict is REJECT:
-   - Fix the issues (delegate via \`task()\` with \`session_id\`)
+   - Fix the issues (delegate via \`task()\` with \`task_id\`)
   - Re-run the rejecting reviewer
   - Repeat until ALL verdicts are APPROVE
 3. Mark \`pass-final-wave\` todo as \`completed\`
@@ -154,7 +154,7 @@ Answer THREE questions:
 ALL three must be YES. "Probably" = NO. "I think so" = NO.

 - **All 3 YES** → Proceed.
- **Any NO** → Reject: resume session with \`session_id\`, fix the specific issue.
+- **Any NO** → Reject: resume with \`task_id\`, fix the specific issue.

 **After gate passes:** Check boulder state:
 \`\`\`
@@ -185,7 +185,7 @@ Final-wave reviewers can finish in parallel before you update the plan file, so

 1. Execute all Final Wave tasks in parallel
 2. If ANY verdict is REJECT:
-   - Fix the issues (delegate via \`task()\` with \`session_id\`)
+   - Fix the issues (delegate via \`task()\` with \`task_id\`)
   - Re-run the rejecting reviewer
   - Repeat until ALL verdicts are APPROVE
 3. Mark \`pass-final-wave\` todo as \`completed\`
@@ -242,7 +242,7 @@ Subagents CLAIM "done" when:

 **Phase 3 is NOT optional for user-facing changes.**
 **Phase 4 gate: ALL three questions must be YES. "Unsure" = NO.**
-**On failure: Resume with \`session_id\` and the SPECIFIC failure.**
+**On failure: Resume with \`task_id\` and the SPECIFIC failure.**
 </verification_rules>`

 export const GEMINI_ATLAS_BOUNDARIES = `<boundaries>
@@ -272,7 +272,7 @@ export const GEMINI_ATLAS_CRITICAL_RULES = `<critical_rules>
 - Send prompts under 30 lines
 - Skip scanned-file lsp_diagnostics (use 'filePath=".", extension=".ts"' for TypeScript projects; directory scans are capped at 50 files)
 - Batch multiple tasks in one delegation
- Start fresh session for failures (use session_id)
+- Start fresh session for failures (do NOT do this; use task_id)

 **ALWAYS**:
 - Include ALL 6 sections in delegation prompts
@@ -280,6 +280,6 @@ export const GEMINI_ATLAS_CRITICAL_RULES = `<critical_rules>
 - Run scanned-file QA after every delegation
 - Pass inherited wisdom to every subagent
 - Parallelize independent tasks
- Store and reuse session_id for retries
+- Store and reuse task_id for retries
 - **USE TOOL CALLS for verification - not internal reasoning**
 </critical_rules>`
@@ -158,7 +158,7 @@ Before moving to the next task, answer these THREE questions honestly:
 3. **Am I confident this doesn't break existing functionality?** (If no → run broader tests)

 - **All 3 YES** → Proceed: mark task complete, move to next.
- **Any NO** → Reject: resume session with \`session_id\`, fix the specific issue.
+- **Any NO** → Reject: resume with \`task_id\`, fix the specific issue.
 - **Unsure on any** → Reject: "unsure" = "no". Investigate until you have a definitive answer.

 **After gate passes:** Check boulder state:
@@ -190,7 +190,7 @@ Final-wave reviewers can finish in parallel before you update the plan file, so

 1. Execute all Final Wave tasks in parallel
 2. If ANY verdict is REJECT:
-   - Fix the issues (delegate via \`task()\` with \`session_id\`)
+   - Fix the issues (delegate via \`task()\` with \`task_id\`)
   - Re-run the rejecting reviewer
   - Repeat until ALL verdicts are APPROVE
 3. Mark \`pass-final-wave\` todo as \`completed\`
@@ -248,7 +248,7 @@ Your job is to CATCH THEM. Assume every claim is false until YOU personally veri

 **Phase 4 gate:** ALL three questions must be YES to proceed. "Unsure" = NO. Investigate until certain.

-**On failure at any phase:** Resume with \`session_id\` and the SPECIFIC failure. Do not start fresh.
+**On failure at any phase:** Resume with \`task_id\` and the SPECIFIC failure. Do not start fresh.
 </verification_rules>`

 export const GPT_ATLAS_BOUNDARIES = `<boundaries>
@@ -276,7 +276,7 @@ export const GPT_ATLAS_CRITICAL_RULES = `<critical_rules>
 - Send prompts under 30 lines
 - Skip scanned-file lsp_diagnostics (use 'filePath=".", extension=".ts"' for TypeScript projects; directory scans are capped at 50 files)
 - Batch multiple tasks in one delegation
- Start fresh session for failures (use session_id)
+- Start fresh session for failures (do NOT do this; use task_id)

 **ALWAYS**:
 - Include ALL 6 sections in delegation prompts
@@ -284,5 +284,5 @@ export const GPT_ATLAS_CRITICAL_RULES = `<critical_rules>
 - Run scanned-file QA after every delegation
 - Pass inherited wisdom to every subagent
 - Parallelize independent tasks
- Store and reuse session_id for retries
+- Store and reuse task_id for retries
 </critical_rules>`
@@ -407,7 +407,7 @@ After delegation, ALWAYS verify: works as expected? follows codebase pattern? MU

 ### Session Continuity

-Every \`task()\` output includes a session_id. **USE IT for follow-ups.**
+Every \`task()\` output includes a task_id. **USE IT for follow-ups.**

 - **Task failed/incomplete** - \`task_id="{id}", prompt="Fix: {error}"\`
 - **Follow-up on result** - \`task_id="{id}", prompt="Also: {question}"\`