Tool: Read
Parameters:
  file_path: skills/vibe/references/vibe-coding.md

# Check if in git repo
git rev-parse --git-dir 2>/dev/null || echo "NOT_GIT"

# Check if ao CLI is available
if command -v ao &> /dev/null; then
  # Search for prior validation failures on similar code
  ao search "validation failures" --limit 5 2>/dev/null || true

  # Check for known anti-patterns from learnings
  ao anti-patterns 2>/dev/null | head -20 || true
else
  # ao not available - skip knowledge injection
  echo "Note: ao CLI not available, skipping knowledge injection"
fi

./scripts/toolchain-validate.sh --gate 2>&1 | tee .agents/tooling/vibe-run.log
TOOL_EXIT=$?

Report to user:

  Grade: F (tools failed)

  Toolchain found CRITICAL issues that must be fixed:
  - See .agents/tooling/<tool>.txt for details

  Fix these issues before re-running /vibe.
  Do NOT generate a false "Grade: B" when tools are failing.

# Check staged changes
git diff --cached --name-only 2>/dev/null | head -10

# Check unstaged changes
git diff --name-only 2>/dev/null | head -10

# Check recent commits
git log --oneline -5 --since="24 hours ago" 2>/dev/null

STOP and return:
  Grade: PASS
  Reason: "No changes detected to review"
  Action: None required

# For "recent" target
git diff --name-only HEAD~3 2>/dev/null | head -20

# For specific path
ls -la <path>

cat .agents/tooling/semgrep.txt
cat .agents/tooling/gitleaks.txt
cat .agents/tooling/gosec.txt
cat .agents/tooling/ruff.txt
cat .agents/tooling/golangci-lint.txt
cat .agents/tooling/shellcheck.txt
cat .agents/tooling/radon.txt
cat .agents/tooling/hadolint.txt

Tool: Task (ALL 3 IN PARALLEL)
Parameters:
  subagent_type: "agentops:security-reviewer"
  model: "haiku"
  description: "Triage security tool findings"
  prompt: |
    TOOL FINDINGS TO TRIAGE:

    ## Gitleaks Output:
    <paste .agents/tooling/gitleaks.txt>

    ## Semgrep Output:
    <paste .agents/tooling/semgrep.txt>

    ## Gosec Output:
    <paste .agents/tooling/gosec.txt>

    For EACH finding, determine verdict:

    TRUE_POSITIVE if:
    - File path exists (not in comments/examples)
    - Not in test fixtures (*/test/*, */mock/*, *_test.go)
    - Not already suppressed (.gitleaksignore, //nolint, # nosec)
    - Pattern matches real credential (not placeholder like "xxx")

    FALSE_POSITIVE if:
    - In test fixtures or examples
    - Already in ignore file
    - Placeholder value (contains "example", "test", "xxx")
    - Dead code path (function never called)

    OUTPUT FORMAT:
    | File:Line | Tool | Finding | Verdict | Reason | Fix (if TRUE_POS) |
    |-----------|------|---------|---------|--------|-------------------|

Tool: Task
Parameters:
  subagent_type: "agentops:code-reviewer"
  model: "haiku"
  description: "Triage linter findings"
  prompt: |
    LINTER FINDINGS TO TRIAGE:

    ## Ruff/Golangci-lint Output:
    <paste .agents/tooling/ruff.txt or golangci-lint.txt>

    ## Shellcheck Output:
    <paste .agents/tooling/shellcheck.txt>

    For EACH finding, apply severity rules:

    FIX_NOW if:
    - Blocks functionality (import error, syntax error)
    - Security implication (bare except, eval usage)
    - Cyclomatic complexity >15 in changed code

    TECH_DEBT if:
    - Style only (line length 81-100)
    - Complexity 10-15
    - Has TODO with issue reference

    NOISE if:
    - Already passing CI
    - No functional impact
    - In generated code

    OUTPUT FORMAT:
    | File:Line | Finding | Priority | Reason |
    |-----------|---------|----------|--------|

Tool: Task
Parameters:
  subagent_type: "agentops:architecture-expert"
  model: "haiku"
  description: "Triage complexity findings"
  prompt: |
    COMPLEXITY FINDINGS TO TRIAGE:

    ## Radon Output:
    <paste .agents/tooling/radon.txt>

    ## Hadolint Output:
    <paste .agents/tooling/hadolint.txt>

    For EACH high-complexity function:
    - Is it in changed files? (only review what's new)
    - Can it be split? (identify extraction points)
    - Is complexity justified? (state machines, parsers OK)

    OUTPUT FORMAT:
    | File:Function | Complexity | In Changed? | Recommendation |
    |---------------|------------|-------------|----------------|

# Vibe Report: <Target>

**Date:** YYYY-MM-DD
**Files Reviewed:** <count>
**Grade:** <A-F>

## Summary
<Overall assessment in 2-3 sentences>

## Gate Decision
[ ] PASS - 0 critical findings
[ ] BLOCK - <count> critical findings must be fixed

## Findings

### CRITICAL
1. **<File:Line>** - <Issue>
   - **Risk:** <what could happen>
   - **Fix:** <how to fix>

### HIGH
1. **<File:Line>** - <Issue>
   - **Fix:** <how to fix>

### MEDIUM
- <File:Line>: <brief issue>

## Aspects Summary
| Aspect | Status |
|--------|--------|
| Semantic | <OK/Issues> |
| Security | <OK/Issues> |
| Quality | <OK/Issues> |
| Architecture | <OK/Issues> |
| Complexity | <OK/Issues> |
| Performance | <OK/Issues> |
| Slop | <OK/Issues> |
| Accessibility | <OK/N/A> |

# Check if ao CLI is available
if command -v ao &> /dev/null; then
  # If CRITICAL findings were discovered, record them as learnings
  if [ "<grade>" = "D" ] || [ "<grade>" = "F" ]; then
    ao memory_store \
      --content "Validation found CRITICAL issues in <target>: <summary of critical findings>" \
      --memory_type "episode" \
      --tags '["validation", "critical", "<area>"]' \
      --source "vibe skill" 2>/dev/null || true
  fi

  # Record any new anti-patterns discovered
  # (Only if a pattern was found that wasn't already known)
  # ao forge transcript <session-log> 2>/dev/null || true
else
  # ao not available - skip result recording
  echo "Note: ao CLI not available, skipping result recording"
fi

# Run prescan for secret detection
./scripts/prescan.sh <target-path>

Before reading any documents, generate a checklist:
- List every entity that should be consistent (agents, states, message types, timeouts)
- List every cross-reference to verify (A mentions B → B exists and matches)
- List every enum/constant that appears in multiple places

| Entity | Doc A (line) | Doc B (line) | Match? |
|--------|--------------|--------------|--------|
| HARVEST_REQUEST sender | auth matrix:152 "Moirai" | comm matrix:1123 "Athena" | ❌ CONFLICT |

## Findings
[List what was found]

## Verification Status
⚠️ Claude-based validation has known limitations.
Recommend external verification with Codex or mechanical diff tools.

## Cross-Reference Tables
[Show the tables built in Step S2]

Tool: Task (ALL IN PARALLEL)
Parameters:
  subagent_type: "agentops:plan-compliance-expert"
  description: "Check spec compliance"
  prompt: "Verify these specs are internally consistent: <file-list>"

Tool: Task
Parameters:
  subagent_type: "agentops:gap-identifier"
  description: "Find spec gaps"
  prompt: "Find missing definitions or broken references in: <file-list>"

Tool: Task
Parameters:
  subagent_type: "agentops:assumption-challenger"
  description: "Challenge assumptions"
  prompt: "Challenge assumptions and find conflicts in: <file-list>"