Audit

Skill senda-labs/DQIII8/.claude/skills/audit

Run complete system health audit of DQIII8 — checks DB integrity, agent performance, pipeline connections, error log, and services. Produces a scored Markdown report.From its SKILL.md

Install

npx -y skills add senda-labs/DQIII8 --skill audit

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

11 stars11 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.5 KB, as published. Nobody here has run it

/audit -- System Health Audit

Triggers the auditor agent to analyze database/dqiii8.db and produce a structured health report.

Usage

/audit
/audit --period 30d       # analyze last 30 days instead of default 7
/audit --agent python-specialist   # scope to one agent

Scope note — `sessions` / `morning_report` / `loop_effectiveness`

.claude/hooks/stop.py:439 writes sessions from every CLI session (INSERT ... ON CONFLICT(session_id) DO UPDATE), gated on _total_actions > 0 (stop.py:436). So sessions is only populated when agent_actions has ≥1 row for that session, which makes a near-empty sessions table a second, independent detector for an agent_actions outage — a real signal to chase, not noise to dismiss. Only morning_report is genuinely bot-only (written solely by bin/ui/dqiii8_bot.py).

loop_effectiveness is a VIEW over objectives, which has 0 rows because the autonomous-loop execution flow (bin/director.py loop mode) isn't in active use yet — an empty result there is still expected, not a symptom to chase.

What it does

Queries all metric tables: agent_actions, error_log, sessions, skill_metrics
Uses views agent_performance and error_keywords_freq
Computes an overall health score (0-100)
Writes a Markdown report to database/audit_reports/audit-YYYY-MM-DD-HH.md
Inserts a summary row in the audit_reports table
Prints a one-line summary to the terminal

Output

[AUDIT] Score: 87/100 | Actions: 106 | Success: 100.0% | Failures: 0 | Unresolved errors: 0
Report: database/audit_reports/audit-2026-03-11-14.md

Score interpretation

Score	Status	Cadencia recomendada
>= 85	HEALTHY	next audit in 7 days
70-85	WARNING	next audit in 3 days
< 70	CRITICAL	next audit in 1 day, notify user

Auto-trigger

The stop.py hook automatically triggers /audit when 7+ days have passed since the last report in audit_reports. Also auto-invoked when errors accumulate during a session.

Agent

Handled by: .claude/agents/auditor.md Model: claude-sonnet-5