metrics.validate-csv
Type: commands | Track: Enterprise
Cross-validate all CSV files in docs/static/data/ against actual framework state. Checks counts, assignee names, and data freshness. Delegates to qa-engineer.
Quick Start
Invoke via ADLC command interface: /metrics:validate-csv
Validation Checks
| Check | Source | Target CSV |
|---|---|---|
| Agent count | .claude/agents/*.md file count | components.csv Agents row |
| Command count | .claude/commands/**/*.md recursive | components.csv Commands row |
| DORA freshness | CSV updated field vs current date | dora.csv — flag if >7 days stale |
| Story IDs unique | stories.csv ID column | No duplicate story IDs |
| Story states valid | stories.csv status column | Only known states allowed |
| Assignee names | stories.csv assignee column | Names match agent registry |
Benefits
- Catches drift between actual framework state and CSV SSOT — prevents
NO_ESTIMATED_COUNTSanti-pattern in retrospectives and standups - Validates DORA freshness — stale DORA data produces inaccurate ceremony outputs
- Prevents
THIN_STORY_INFLATION— validates story records have required fields (AC, KR mapping, points) - Single command replaces manual cross-referencing across 6+ CSV columns
When to Use
| Attribute | Detail |
|---|---|
| Persona | QA Engineer |
| Trigger | Data integrity check — before a sprint ceremony (standup, review, retro) to ensure CSV data is accurate, or after a sprint planning session when multiple stories and components were added |
| Business Value | Cross-validates CSV data vs framework state — prevents ceremonies from reporting inaccurate metrics (e.g., component counts that are off by 3, stale DORA values from last sprint); evidence of data integrity for governance audits |
| Frequency | Sprint |
Example: As a QA Engineer, I need to validate CSV data before the sprint review because the review template reads from stories.csv and components.csv, and any drift between actual state and CSV values will produce an inaccurate stakeholder report. I run /metrics:validate-csv which cross-validates agent/command/skill counts against actual directory scans, checks DORA freshness, and validates story state transitions, producing a PASS/FAIL report with specific drift items to remediate.
Delegates to qa-engineer. Reads CSV and framework files only — no mutations.