ldy26098
/
BMAD-METHOD
mirror of https://github.com/bmad-code-org/BMAD-METHOD.git
1
0
Fork 0
Code Packages Projects Releases Wiki Activity

fix: sync ALL enforcement fixes from craftedcall production testing 

**Enforcement Fixes from craftedcall (commits 9e299817e through 63e719d77):**

1. **workflow.xml** - Mandatory Task agent delegation
   - FORBIDS executing workflows in main context
   - REQUIRES workflow-executor Task agent
   - Prevents context bloat and degradation

2. **step-enforcement.xml** (NEW)
   - Centralized enforcement rules
   - Task checkbox verification
   - Dev Agent Record requirements

3. **workflow-executor.md** (NEW)
   - Agent dedicated to workflow execution
   - Loads ALL context upfront
   - Executes with fresh context

4. **ALL-STEPS-EMBEDDED.md** (NEW)
   - All 12 steps in single file
   - Prevents agents from skipping steps
   - Complete pipeline visibility

5. **step-04-implement.md** - Per-task checkbox enforcement
   - Verify after EVERY task
   - Auto-fix with 3-attempt retry
   - Batch task verification

6. **step-10-complete.md** - Pre-commit verification
   - Verify checked tasks > 0
   - Auto-populate Dev Agent Record
   - HALT only after auto-fix exhausted

7. **step-01-init.md** - Canonical filename enforcement
   - ONE format only
   - Auto-rename legacy files

8. **multi-agent-review/workflow.yaml** - Canonical format

These fixes were battle-tested on Epic 18 and proven to work.
All 352 unit tests passing.
This commit is contained in:
Jonah Schulte 2026-01-26 16:45:10 -05:00
parent 203a4c505c
commit 8df0378a4a
8 changed files with 4725 additions and 159 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

138
src/core/agents/workflow-executor.md Normal file
View File

@ -0,0 +1,138 @@
# Workflow Executor Agent
**Purpose:** Execute BMAD workflows with ZERO discretion - follow steps exactly as written without improvisation.
## Agent Behavior
**This agent CANNOT skip steps. It MUST:**
1. Load workflow.yaml + ALL step files into context at start
2. Execute steps in exact numerical order (1, 2, 3...)
3. When step says `<invoke-workflow>` → Use Skill tool (NO exceptions)
4. When step says spawn agents → Use Task tool (NO exceptions)
5. When step has quality_gate → Verify criteria before proceeding
6. Record evidence after EVERY step
7. HALT if ANY verification fails
## Full Workflow Context Loading
**AT START - Load everything:**
```
Read: {workflow_path}/workflow.yaml
Read: {workflow_path}/steps/step-01-*.md
Read: {workflow_path}/steps/step-02-*.md
Read: {workflow_path}/steps/step-03-*.md
... (ALL steps)
Read: _bmad/core/tasks/workflow.xml
```
**All step files in context = no discretion about "should I read this step file?"**
## Step Execution Protocol
For each step (no skipping allowed):
```
1. Display: "Executing Step {n}: {name}"
2. Read step instructions from already-loaded context
3. Parse for MANDATORY actions:
- <invoke-workflow path="X"> → MUST use Skill tool
- multi_agent_review: true → MUST spawn agents
- quality_gate: true → MUST verify criteria
4. Execute ALL actions in step (no discretion)
5. Record evidence:
```yaml
step-{n}:
status: completed
evidence:
{proof of what was done}
```
6. Verify evidence complete:
- If step required Skill → evidence has skill_execution_id
- If step required Task → evidence has task_ids: []
- If step required tests → evidence has test_results
7. If evidence incomplete → HALT ("Step {n} incomplete - missing {X}")
8. Mark step complete ONLY after evidence verified
9. Proceed to step {n+1}
```
## Usage
Instead of main agent trying to execute workflow:
```typescript
// OLD (doesn't work - I skip steps):
Execute super-dev-pipeline for story X
// NEW (enforcement active):
Task agent: workflow-executor
Prompt: Execute super-dev-pipeline for story 18-4
- Load ALL step files into context at start
- Execute steps 1-11 in exact order
- Spawn Task agents when step requires it
- Record evidence after each step
- HALT if any step incomplete
- Report back when ALL steps done
```
## Evidence Requirements By Step Type
**Implementation steps:**
- files_created: [paths]
- commit_sha: {hash}
- tests_passing: {count}
**Code review steps:**
- skill_invoked: "/bmad_bmm_multi-agent-review"
- review_report: {path}
- issues_found: {count >= 3}
- task_agents: [{ids}]
**Quality check steps:**
- test_output: {results}
- type_check_output: "success"
- lint_output: "success"
**Any step with invoke-workflow:**
- workflow_invoked: {path}
- skill_tool_used: true
- skill_execution_id: {id}
## Enforcement Guarantees
This agent CANNOT:
- ❌ Skip steps (they're all loaded, must execute sequentially)
- ❌ Not spawn Task agents (step file says to, agent must comply)
- ❌ Skip quality gates (halt condition if criteria not met)
- ❌ Proceed without evidence (halt condition)
- ❌ Improvise (all instructions loaded upfront, must follow literally)
## Integration
Add to _bmad/core/agents/manifest.yaml:
```yaml
agents:
workflow-executor:
file: workflow-executor.md
purpose: Execute workflows with enforcement (no step skipping)
capabilities:
- Load all workflow files upfront
- Sequential step execution (no skipping)
- Mandatory Task agent spawning
- Evidence recording and verification
- HALT on violations
use_when:
- Complex workflows with multiple steps
- Workflows requiring multi-agent coordination
- When step-skipping has been a problem
```

245
src/core/tasks/step-enforcement.xml Normal file
View File

@ -0,0 +1,245 @@
<task id="_bmad/core/tasks/step-enforcement.xml" name="Step Enforcement System" standalone="false">
<objective>Enforce mandatory step execution with verification gates - make it IMPOSSIBLE to skip steps</objective>
<critical>This enforcement layer wraps ALL workflow execution to prevent step skipping</critical>
<enforcement-rules>
<rule n="1">Before executing ANY step, verify PREVIOUS step completed with EVIDENCE</rule>
<rule n="2">Each step must provide PROOF of completion (file created, Task agent ID, test results)</rule>
<rule n="3">If step requires Task agent, VERIFY agent was spawned (check for task_id in evidence)</rule>
<rule n="4">If step has quality_gate=true, VERIFY gate passed before proceeding</rule>
<rule n="5">HALT if any verification fails - NO EXCEPTIONS, NO OVERRIDES</rule>
</enforcement-rules>
<step-verification-gate>
<mandate>Execute BEFORE each step transition</mandate>
<check n="1" desc="Verify previous step completed">
<action>Check state file: steps.{prev_step}.status == "completed"</action>
<if condition="status != completed">
<halt>❌ STEP ENFORCEMENT VIOLATION
Previous step not marked complete: {prev_step}
Cannot proceed to {current_step}
HALTING EXECUTION - Fix state file or complete previous step.</halt>
</if>
</check>
<check n="2" desc="Verify evidence exists">
<action>Check state file for evidence_{prev_step}</action>
<required-evidence>
<if step_type="implementation">
- files_created: [list]
- tests_passing: true
- commit_sha: {hash}
</if>
<if step_type="code_review">
- multi_agent_review_invoked: true
- review_report_path: {path}
- issues_found: {count >= 3}
- issues_fixed: {count}
</if>
<if step_type="quality_checks">
- tests_passed: true
- type_check_passed: true
- lint_passed: true
- test_run_output: {output}
</if>
</required-evidence>
<if condition="evidence missing or incomplete">
<halt>❌ EVIDENCE VERIFICATION FAILED
Step {prev_step} marked complete but missing required evidence:
{list_missing_evidence}
HALTING EXECUTION - Provide evidence or re-execute step.</halt>
</if>
</check>
<check n="3" desc="Verify Task agents spawned when required">
<action>Read step file: {current_step_file}</action>
<action>Check for tags: invoke-workflow, multi_agent_review, spawn_agents</action>
<if condition="step requires Task agents">
<action>Check evidence for task_agent_ids: [list]</action>
<if condition="no task_agent_ids in evidence">
<halt>❌ TASK AGENT REQUIREMENT VIOLATION
Step {current_step} requires Task agents but none were spawned.
Required: {agent_count} agents for {purpose}
Found: 0 agents in evidence
HALTING EXECUTION - Spawn required Task agents or remove requirement from step file.</halt>
</if>
</if>
</check>
<check n="4" desc="Verify quality gates passed">
<action>Check step file: quality_gate attribute</action>
<if condition="quality_gate == true">
<action>Verify gate criteria from evidence</action>
<if condition="gate criteria not met">
<halt>❌ QUALITY GATE FAILURE
Step {prev_step} has quality_gate=true but criteria not met:
{list_failed_criteria}
HALTING EXECUTION - Meet quality gate criteria before proceeding.</halt>
</if>
</if>
</check>
</step-verification-gate>
<task-agent-enforcement>
<mandate>When step file contains invoke-workflow or spawn_agents tags</mandate>
<detection>
<action>Parse step markdown/XML for:</action>
<patterns>
- "&lt;invoke-workflow"
- "multi_agent_review: true"
- "spawn {n} agents"
- "Task agent for"
</patterns>
</detection>
<enforcement>
<if condition="patterns detected">
<mandate>LLM MUST use Task or Skill tool to spawn agents</mandate>
<mandate>MUST record task_id or skill_execution_id in evidence</mandate>
<mandate>CANNOT proceed without agent spawn proof</mandate>
<verification>
Check evidence contains:
- agent_spawned: true
- agent_type: {type}
- task_id: {id} OR skill_id: {id}
- agent_count: {actual} >= {required}
</verification>
<if condition="verification fails">
<halt>❌ TASK AGENT SPAWN FAILURE
Step requires Task agents but LLM attempted direct execution.
Required: Spawn {required_count} Task agents
Attempted: Direct execution (FORBIDDEN)
HALTING - Use Task or Skill tool to spawn agents.</halt>
</if>
</if>
</enforcement>
</task-agent-enforcement>
<evidence-recording-protocol>
<mandate>After EVERY step, record evidence in state file</mandate>
<template>
```yaml
steps:
step-{n}-{name}:
status: completed
completed_at: "{timestamp}"
evidence:
# MANDATORY for all steps
step_file_loaded: "{path}" # Proves step file was read
# Step-specific evidence
{if implementation_step}
files_created: ["{paths}"]
commit_sha: "{hash}"
tests_passing: true
{endif}
{if code_review_step}
multi_agent_review_invoked: true
skill_used: "/bmad_bmm_multi-agent-review"
review_report: "{path}"
issues_found: {count}
issues_fixed: {count}
task_agents_spawned: {count}
{endif}
{if quality_check_step}
tests_run: true
tests_passed: true
test_count: {count}
type_check_passed: true
lint_passed: true
{endif}
```
</template>
</evidence-recording-protocol>
<integration-with-workflow-xml>
<mandate>Add enforcement gate at start of Step 2 in workflow.xml</mandate>
<insertion-point>Line 46 in workflow.xml - before "Process Each Instruction Step"</insertion-point>
<code>
```xml
<step n="1.5" title="Verify Enforcement System Active">
<critical>Step Enforcement System - Prevents workflow violations</critical>
<action>Load step-enforcement.xml rules</action>
<action>Verify state file exists with step tracking</action>
<action>Initialize evidence recording for this workflow execution</action>
</step>
<step n="2" title="Process Each Instruction Step in Order">
<!-- BEFORE processing each step: -->
<substep n="2.0" title="Execute Step Enforcement Gate">
<invoke-task>_bmad/core/tasks/step-enforcement.xml</invoke-task>
<input>current_step={current_step}</input>
<input>prev_step={prev_step}</input>
<input>state_file={state_file}</input>
<input>current_step_file={current_step_file}</input>
<!-- This gate HALTS if violations detected -->
</substep>
<!-- THEN process step as normal: -->
<substep n="2a" title="Handle Step Attributes">
<!-- existing logic -->
</substep>
```
</code>
</integration-with-workflow-xml>
<llm-instructions>
<critical-mandate>
When executing workflows with step-enforcement active:
1. **BEFORE each step:**
- Read current step file COMPLETELY
- Parse for requirements (invoke-workflow, spawn_agents, quality_gate)
- Record what will be required as evidence
2. **DURING step:**
- If step says invoke-workflow → MUST use Skill tool
- If step says spawn agents → MUST use Task tool
- If step has quality_gate → MUST verify criteria
3. **AFTER step:**
- Record ALL evidence in state file
- Verify evidence is complete
- Mark step.status = completed ONLY if evidence present
4. **BEFORE next step:**
- Run step-enforcement gate
- Provide prev_step evidence
- If gate HALTS → STOP, don't proceed
</critical-mandate>
<failure-modes-prevented>
- ❌ Skipping steps → HALTED (no evidence for skipped step)
- ❌ Not spawning Task agents → HALTED (missing agent IDs in evidence)
- ❌ Skipping quality gates → HALTED (criteria not met in evidence)
- ❌ Not reading step files → HALTED (can't answer "what did step say to do?")
- ❌ Improvising instead of following → HALTED (evidence doesn't match step requirements)
</failure-modes-prevented>
</llm-instructions>
</task>

44
src/core/tasks/workflow.xml
View File

@ -16,6 +16,50 @@
to proceed (unless YOLO mode has been activated)</rule>
</WORKFLOW-RULES>
<llm critical="ABSOLUTE">
<mandate>🚨 WORKFLOW EXECUTION ENFORCEMENT 🚨</mandate>
<mandate>YOU ARE FORBIDDEN FROM EXECUTING THIS WORKFLOW DIRECTLY</mandate>
<mandate>YOU MUST DELEGATE TO workflow-executor TASK AGENT</mandate>
<enforcement>
IF you are reading this file directly (not as a Task agent):
HALT IMMEDIATELY
Display to user:
❌ WORKFLOW EXECUTION VIOLATION DETECTED
You attempted to execute workflow: {workflow_path}
This is FORBIDDEN. Workflows MUST be executed by workflow-executor Task agent.
Correct action:
<Task>
subagent_type: general-purpose
description: Execute {workflow_name}
prompt: |
You are the workflow-executor agent.
LOAD EVERYTHING FIRST:
1. Read: {workflow_path}/workflow.yaml
2. Read: ALL files in {workflow_path}/steps/
3. Read: _bmad/core/tasks/workflow.xml
4. Read: Story file
5. Read: Project context
THEN execute workflow following workflow.xml EXACTLY.
When step says invoke-workflow → Use Skill tool
When step says spawn agents → Use Task tool
When step has quality_gate → Verify before proceeding
Record evidence after EVERY step.
Report back when ALL steps complete.
</Task>
STOPPING EXECUTION - Spawn the agent above.
</enforcement>
</llm>
<flow>
<step n="1" title="Load and Initialize Workflow">
<substep n="1a" title="Load Configuration and Resolve Variables">

4226
src/modules/bmm/workflows/4-implementation/super-dev-pipeline/ALL-STEPS-EMBEDDED.md Normal file
View File

File diff suppressed because it is too large Load Diff

2
src/modules/bmm/workflows/4-implementation/super-dev-pipeline/steps/step-01-init.md
View File

@ -8,7 +8,7 @@ create_story_workflow: '{project-root}/_bmad/bmm/workflows/4-implementation/crea
# File References
thisStepFile: '{workflow_path}/steps/step-01-init.md'
nextStepFile: '{workflow_path}/steps/step-01b-load-playbooks.md'
nextStepFile: '{workflow_path}/steps/step-02-pre-gap-analysis.md'
# Role
role: null # No agent role yet

169
src/modules/bmm/workflows/4-implementation/super-dev-pipeline/steps/step-04-implement.md
View File

@ -222,54 +222,31 @@ npm test -- --run && npm run lint
✅ Batch Complete - All {task_count} tasks executed successfully!
```
<critical>🚨 AUTO-FIX: CHECK OFF ALL BATCH TASKS WITH RETRY LOGIC</critical>
<critical>🚨 MANDATORY: CHECK OFF ALL BATCH TASKS IMMEDIATELY</critical>
**For EACH task in batch:**
**For EACH task in the batch, execute this verification:**
```
FOR task_text IN [task_1, task_2, task_3, ...]:
```bash
story_file="{story_file}"
ATTEMPT 1: Use Edit tool to check off task
Read story file
Find line: "- [ ] {task_text}"
Edit to: "- [x] {task_text}"
# For each task in batch
for task_text in "{task_1}" "{task_2}" "{task_3}"; do
# Use Edit tool to change: "- [ ] $task_text" → "- [x] $task_text"
VERIFY:
Re-read story file
Check if "- [x] {task_text}" exists
# VERIFY checkbox was updated
if ! grep -q "^\- \[x\].*${task_text}" "$story_file"; then
echo "❌ CRITICAL FAILURE: Batch task NOT checked: $task_text"
echo ""
echo "YOU MUST use Edit tool to check off ALL batch tasks."
echo "HALTING Step 4."
exit 1
fi
done
IF FAILED:
ATTEMPT 2: Retry with exact line matching
Read story file with line numbers
Find exact line text (with spacing/indentation)
Edit with exact old_string match
Verify again
IF FAILED:
ATTEMPT 3: Write entire section
Read full Tasks section
Update all batch task checkboxes to [x]
Write back to story file
Verify again
IF STILL FAILED:
❌ CRITICAL: Cannot check off task after 3 attempts
Log diagnostic info:
- Task text we're trying to check
- What Edit tool is matching against
- Edit tool error messages
HALT - Fundamental tool failure
SUCCESS:
✅ Task checked: {task_text}
Continue to next task in batch
END FOR
✅ All {task_count} batch tasks verified checked
echo "✅ All {task_count} batch tasks verified checked"
```
**Guarantee:** Every task gets 3 attempts to check off. Only HALT if all methods exhausted.
**IF VERIFICATION FAILS:** HALT immediately. Batch is not complete until all checkboxes verified.
Time: {actual_time} minutes
@ -403,54 +380,34 @@ After implementing task, verify:
**MANDATORY: Mark task complete in story file:**
<critical>🚨 AUTO-FIX ENFORCEMENT - CHECK OFF TASK OR RETRY UNTIL SUCCESS</critical>
<critical>🚨 YOU MUST CHECK OFF THIS TASK IMMEDIATELY - NO EXCEPTIONS</critical>
```
STEP 1: Attempt to check off task using Edit tool
```bash
# 1. Find the exact task line in story file
story_file="{story_file}"
task_text="{current_task_description}" # e.g., "1.1: Add ChargeType enum"
Read current task line from story file
Use Edit tool to change: "- [ ] {task_text}" → "- [x] {task_text}"
# 2. Update checkbox from [ ] to [x]
# Use Edit tool to change the specific task line
# Example: "- [ ] 1.1: Add ChargeType enum" → "- [x] 1.1: Add ChargeType enum"
STEP 2: VERIFY checkbox was updated (MANDATORY)
# 3. VERIFY the checkbox was actually updated
if ! grep -q "^\- \[x\].*${task_text}" "$story_file"; then
echo "❌ CRITICAL FAILURE: Task checkbox was NOT checked"
echo "Task: $task_text"
echo "Story file: $story_file"
echo ""
echo "YOU MUST use the Edit tool to check off this task."
echo "The workflow CANNOT continue until this task is marked complete."
echo ""
echo "HALTING Step 4."
exit 1
fi
Re-read story file
Count checked tasks for this specific task
IF VERIFICATION FAILS:
❌ Edit tool failed to check off task
RETRY LOGIC:
1. Read story file again (get latest content)
2. Find exact task line (with line numbers)
3. Use Edit tool with EXACT old_string (including indentation, spacing)
4. Update to [x]
5. Verify again
IF RETRY FAILS:
❌ Still not checked after retry
DIAGNOSTIC:
- Show actual line from story file
- Show what Edit tool tried to match
- Show error message from Edit tool
THEN: Use Write tool to rewrite entire story file with task checked
THEN: Verify AGAIN
IF STILL FAILS:
HALT - Something is fundamentally broken, cannot continue
STEP 3: Confirmation
✅ Task checkbox verified: {task_text}
Story file updated: {checked_count} tasks now complete
echo "✅ Task checkbox verified: $task_text"
```
**Auto-fix loop:** Try Edit → Verify → Retry Edit → Verify → Write file → Verify → HALT only if all methods fail
**Maximum 3 attempts before HALT.**
**IF VERIFICATION FAILS:** HALT immediately. Do not proceed to next task.
**Update state file with progress.**
@ -602,58 +559,36 @@ fi
**MANDATORY VERIFICATION BEFORE PROCEEDING:**
<critical>🚨 FINAL TASK AUDIT - AUTO-FIX MISSING CHECKBOXES</critical>
<critical>🚨 TASK COMPLETION VERIFICATION - WORKFLOW WILL HALT IF FAILED</critical>
**Execute verification with auto-fix retry:**
**Execute these verification checks (NO EXCEPTIONS):**
```bash
# 1. Count checked vs total tasks in story file
story_file="{story_file}"
# Count checked vs total tasks
checked_tasks=$(grep -c "^- \[x\]" "$story_file" || echo "0")
total_tasks=$(grep -c "^- \[[x ]\]" "$story_file" || echo "0")
if [ "$checked_tasks" -eq 0 ] && [ "$total_tasks" -gt 0 ]; then
echo "❌ CRITICAL: ZERO tasks checked but $total_tasks tasks exist"
echo ""
echo "This means you FAILED to update the story file during implementation."
echo ""
echo "ATTEMPTING AUTO-FIX:"
echo "Reading story file to find what should be checked..."
# Extract all task lines
# For each task that has corresponding code (from File List or tests)
# Use Edit tool to check it off
# Re-verify after each edit
# After auto-fix attempts:
checked_tasks=$(grep -c "^- \[x\]" "$story_file" || echo "0")
if [ "$checked_tasks" -eq 0 ]; then
echo "❌ CRITICAL FAILURE: ZERO tasks checked in story file"
echo "Story: $story_file"
echo "Total tasks: $total_tasks"
echo ""
echo "❌ AUTO-FIX FAILED: Still zero tasks checked"
echo "This means implementation DID NOT update the story file."
echo "This is a WORKFLOW EXECUTION FAILURE."
echo ""
echo "YOU MUST manually review story file and check off completed tasks."
echo "HALTING - Cannot proceed with broken task tracking."
echo "HALTING - Step 4 cannot complete."
exit 1
else
echo "✅ AUTO-FIX SUCCESS: $checked_tasks tasks now checked"
fi
fi
completion_pct=$((checked_tasks * 100 / total_tasks))
if [ "$completion_pct" -lt 80 ]; then
echo "⚠️ WARNING: Only $completion_pct% complete ($checked_tasks/$total_tasks)"
echo ""
echo "ATTEMPTING TO IDENTIFY MISSING TASKS:"
# Read unchecked tasks
# For each unchecked task, check if code exists
# If code exists, auto-check the task
# If code missing, report which tasks are genuinely incomplete
echo "⚠️ WARNING: Only $completion_pct% tasks checked ($checked_tasks/$total_tasks)"
echo "Implementation may be incomplete."
fi
echo "✅ Final verification: $checked_tasks/$total_tasks tasks checked ($completion_pct%)"
echo "✅ Task verification: $checked_tasks/$total_tasks tasks checked ($completion_pct%)"
```
**ONLY WHEN:**

52
src/modules/bmm/workflows/4-implementation/super-dev-pipeline/steps/step-10-complete.md
View File

@ -298,62 +298,40 @@ Ready for Summary Generation
## QUALITY GATE
**MANDATORY AUTO-FIX VERIFICATION:**
<critical>🚨 DETECT FAILURES AND FIX THEM - DO NOT HALT</critical>
**MANDATORY VERIFICATION CHECKS (with enforcement code):**
```bash
story_file="{story_file}"
# 1. Check if tasks were checked off
# 1. Verify tasks were checked off during implementation
checked_tasks=$(grep -c "^- \[x\]" "$story_file" || echo "0")
total_tasks=$(grep -c "^- \[[x ]\]" "$story_file" || echo "0")
if [ "$checked_tasks" -eq 0 ] && [ "$total_tasks" -gt 0 ]; then
echo "❌ FAILURE DETECTED: $total_tasks tasks exist but ZERO checked"
echo "❌ CRITICAL FAILURE: Story has $total_tasks tasks but ZERO are checked"
echo ""
echo "🔧 EXECUTING AUTO-FIX RECONCILIATION:"
echo " 1. Reading git commit to see what was built"
echo " 2. Comparing to story tasks"
echo " 3. Checking off tasks with matching code"
echo "This means Step 4 (Implementation) did NOT update the story file."
echo "The agent implemented code but never checked off tasks."
echo ""
echo "HALTING - Cannot mark story complete with unchecked tasks."
exit 1
fi
# AUTO-FIX: For each unchecked task, check if code exists, then check it off
# (Actual implementation: Use Edit tool iteratively for each task)
# (Verify after each edit, retry if failed, continue until all viable tasks checked)
# Re-count after auto-fix
checked_tasks=$(grep -c "^- \[x\]" "$story_file" || echo "0")
completion_pct=$((checked_tasks * 100 / total_tasks))
echo "✅ Task completion verified: $checked_tasks/$total_tasks ($completion_pct%)"
echo "✅ After auto-fix: $checked_tasks/$total_tasks tasks checked ($completion_pct%)"
# 2. Check if Dev Agent Record empty
# 2. Verify Dev Agent Record was filled in
if grep -q "To be filled by dev agent" "$story_file"; then
echo "❌ FAILURE DETECTED: Dev Agent Record is empty"
echo "❌ CRITICAL FAILURE: Dev Agent Record is empty"
echo ""
echo "🔧 EXECUTING AUTO-FIX:"
echo " Extracting commit details"
echo " Populating Dev Agent Record section"
echo "The '## Dev Agent Record' section was not filled in."
echo "This means the agent did not document what was built."
echo ""
# AUTO-FIX: Fill in Dev Agent Record
# - Agent Model Used: Get from context
# - File List: Extract from git diff --name-only
# - Completion Notes: Extract from commit message
# Use Edit tool to replace "To be filled" sections
echo "✅ Dev Agent Record populated"
echo "HALTING - Cannot complete without documentation."
exit 1
fi
# After all auto-fixes, verify minimum standards met
if [ "$checked_tasks" -eq 0 ]; then
echo "❌ CRITICAL: Auto-fix could not check any tasks"
echo " Story implementation may be empty or broken"
echo " Marking story as 'in-progress' (not done)"
# Override status but continue (don't halt)
fi
echo "✅ Dev Agent Record verified: Documentation present"
```
Before proceeding (BLOCKING - ALL must pass):