/phase1c-modernization

Phase 1c: Modernization Audit

Detect outdated hook patterns after registration sync. Runs automatically unless `--skip-modernization` is passed.

Static Pattern Check (always runs)

Run the modernization checker script:

python3 scripts/check_hook_modernization.py

For JSON output (programmatic):

python3 scripts/check_hook_modernization.py --json

Patterns Detected

| Pattern | Severity | Description | |---------|----------|-------------| | `invalid-post-decision` | error | PostToolUse using `"ALLOW"` or `"approve"` instead of `"block"` or omitting | | `deprecated-pre-decision` | warning | PreToolUse using `"decision"`/`"reason"` instead of `hookSpecificOutput.permissionDecision` | | `missing-stdin-error-handling` | warning | Hook reads stdin without try/except for JSONDecodeError | | `noisy-no-op` | warning | PostToolUse prints stdout on no-op paths (silent exit is preferred) |

Interpreting Results

• **Errors** block the workflow. Fix before proceeding.
• **Warnings** are informational. Review and fix when convenient.

Fixing Common Issues

**invalid-post-decision**: Replace `{"decision": "ALLOW"}` with `{}` (empty dict). PostToolUse hooks signal "allow" by omitting the decision field entirely, or by producing no stdout.

**deprecated-pre-decision**: Replace top-level `"decision"` and `"reason"` with nested `hookSpecificOutput`:

# Before (deprecated):
{"decision": "approve", "reason": "safe"}

# After (current):
{
    "hookSpecificOutput": {
        "hookEventName": "PreToolUse",
        "permissionDecision": "allow",
        "permissionDecisionReason": "safe",
    }
}

**missing-stdin-error-handling**: Wrap stdin reads:

try:
    data = json.loads(sys.stdin.read())
except (json.JSONDecodeError, ValueError):
    sys.exit(0)

Tome Research Audit (runs by default)

After the static check, dispatch a tome:code-search agent to verify hook patterns against current community practices.

When to Run

• Default: runs on every `/update-plugins` invocation
• Skip with: `--skip-research`
• Useful after Claude Code SDK updates to catch new patterns

Workflow

1. Collect the list of hook event types used in the project 2. Dispatch `tome:code-search` agent with query:

   Claude Code plugin hooks PostToolUse PreToolUse
   hookSpecificOutput best practices 2025 2026

3. Compare findings against local hook implementations 4. Report any modernization gaps not caught by static checks

Agent Dispatch

Agent(tome:code-searcher)
Prompt: Search GitHub for current Claude Code plugin hook
patterns. Focus on PostToolUse and PreToolUse response
formats, hookSpecificOutput usage, and deprecated patterns.
Return structured findings with repo URLs and code snippets.

Output

The agent returns findings in this structure:

{
  "findings": [
    {
      "repo": "owner/repo",
      "pattern": "description of pattern",
      "relevance": "high|medium|low",
      "url": "https://github.com/..."
    }
  ]
}

Cross-reference against local hooks to identify gaps.