frontend-skills-stats

Frontend skills stats

Step 0: Gather context

Run these Bash commands before proceed:

• ls "$(git rev-parse --show-toplevel 2>/dev/null)/.claude/hooks/"*.sh 2>/dev/null | wc -l -- installed hook scripts
• jq '[.hooks[]?[]?.hooks[]?] | length' "$(git rev-parse --show-toplevel 2>/dev/null)/skill-manifest.json" -- wired hooks
• ls ~/.claude/hook-metrics/*.json 2>/dev/null | wc -l -- session summaries
• ls ~/.claude/hook-metrics/*.json 2>/dev/null | head -1 | xargs -I{} jq -r '.date' {} -- earliest date
• ls ~/.claude/hook-metrics/*.json 2>/dev/null | tail -1 | xargs -I{} jq -r '.date' {} -- latest date

Metrics dir: ~/.claude/hook-metrics/

Your task

Analyze frontend-skills harness across all session metrics. Read every JSON in ~/.claude/hook-metrics/. Produce prioritized report.

1. Latency profile

Parse perf_ms from each session summary (added 2.2.2). Each hook:

Hook	P50 (ms)	P95 (ms)	Invocations	Total wall-clock

Flag hooks:

• P95 > 100ms -> perf budget breach
• P95 > 500ms -> critical
• Invocations = 0 all sessions -> zero-fire candidate

2. Rule activity

Each rule fired once+:

Rule	Blocks	Warns	Nudges	Info	Diagnostic

Detect new tier usage (nudge, info, diagnostic -- added 2.2.2). Report adoption rate.

3. Silent hooks (zero fires)

List wired hooks, zero fires all sessions. Prune candidates.

4. Over-aggressive hooks

• Blocks-per-session > 3 -> too strict, demote to warn
• Same rule blocked >= 2x one session -> Claude retrying, message unclear
• block-strict, no escape-hatch adoption -> rule too harsh

5. Under-enforced rules

Cross-ref CLAUDE.md rules vs hook activity:

• Rule in CLAUDE.md, no hook -> advisory; add hook or accept doc-only
• Rule has hook, zero fires -> Claude never violates; safe
• Rule has hook, high fire rate -> document in README prominently

6. Session health signals

• Sessions with hook errors (exit > 0 not in {0, 2}) -> config bug
• Sessions with >20 blocks same rule -> feedback loop
• Sessions with 0 hook fires -> perfect compliance or hooks dead

7. Manifest drift check

Compare skill-manifest.json vs .claude/settings.json and hooks/hooks.json:

bash scripts/generate-hook-configs.sh --check

Drift detected: RED FLAG -- drift bug regression. Run --apply to fix.

Output format

Structured report. Tables for numeric data. End with prioritized action list (max 5 items). If <5 session files, mark recommendations preliminary.

Plain markdown. No emojis.