skills/analytics/SKILL.md
Use when you need to analyze work tracker data for agent metrics or session stats.
npx skillsauth add seokan-jeong/team-shinchan team-shinchan:analyticsInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Analyze work tracker events for observability insights.
/team-shinchan:analytics # Current session summary
/team-shinchan:analytics --report full # Full analytics report
/team-shinchan:analytics --agent bo # Single agent detail
/team-shinchan:analytics --trace trace-xxx # Trace timeline
/team-shinchan:analytics --period 7d # Last 7 days only
| Arg | Default | Description |
|-----|---------|-------------|
| --report full | (session) | Generate full analytics report |
| --agent {name} | (all) | Detail for a single agent |
| --trace {id} | (none) | Timeline for a specific trace ID |
| --period {N}d | (all) | Filter events to the last N days |
| --format table | json | Output as text table instead of JSON |
Use .shinchan-docs/work-tracker.jsonl as the data source.
If file does not exist:
No work tracker log found.
Events will be recorded automatically during Claude Code sessions.
Execute the analytics script via Bash:
node "${CLAUDE_PLUGIN_ROOT}/src/analytics.js" .shinchan-docs/work-tracker.jsonl [options]
Map user arguments to script flags:
--report full → no extra flags (full report is default)--agent <name> → --agent <name>--trace <id> → --trace <id>--period <N>d → --period <N>d--format table--format table or default: display the text table output directlyAfter displaying raw output, provide a brief interpretation:
| Metric | Description | |--------|-------------| | Agent call count | How many times each agent was invoked | | Avg duration | Average time from agent_start to agent_done | | Success rate | Ratio of agent_done to agent_start | | Session stats | Events, file changes, delegations per session | | Event distribution | Percentage breakdown of event types | | Delegation graph | Which agent delegated to which, with counts |
src/analytics.js (Node.js, built-in modules only).shinchan-docs/work-tracker.jsonltrace-init hooktesting
Default-on interview option-quality panel — N diverse generators produce structure-free options, a SelfCheckGPT majority-vote consensus filters hallucinations, a SteerConf cautious-confidence judge scores survivors, and a deterministic top-K is returned. Workflow tier; the single fierce-* skill that is ON by default.
development
Deterministic adversarial code review for high-stakes scope — independent per-dimension review, a non-skippable per-finding refutation, completeness + interaction critics, and a deterministic 3-lens rubric judge panel. Opt-in main-loop Workflow tier.
data-ai
Deterministic loop-until-done for high-stakes long-running tasks — a worker/verifier loop the script bounds by iteration cap, token budget, and stagnation, closed by an Action-Kamen gate. Opt-in main-loop Workflow tier.
testing
Deterministic adversarial debate for high-stakes or irreversible decisions — mandatory refutation plus a scored judge panel. Opt-in main-loop Workflow tier.