---
description: Regenerate the PRD Plugin reports (skill-usage now; request/health if their helpers are installed)
allowed-tools: Bash(python:*)
---

Regenerate the available read-only reports to files, then summarize. Resolve the
scripts under `.prd_plugin/scripts/` (downstream) or `scripts/` (hub). Run only
the ones that exist:

- `skill_usage_report.py --output .prd_plugin/local/skill-usage-report.md`
- `request_report.py --config .prd_plugin/config.json --output .prd_plugin/local/request-report.md` (if present)
- `prd_doctor.py --repo-root . --output .prd_plugin/local/prd-doctor.md` (if present)
- `prd_status.py` (print to chat)

Report which files were written and flag anything notable (stale items, work
recorded without verification-skill use). Do not change project state.

## Optional delegated summary

Keep report generation deterministic. After the files above exist, inspect the
effective policy with `python <scripts>/prd_reporting.py --repo-root . policy`.
For an enabled `report_summary` task, build its source-referenced bundle with
`python <scripts>/prd_reporting.py --repo-root . bundle --task report_summary`
and ask AI-Collab's delegated-reporting executor to summarize that bundle with
the configured `fast-capable` profile. Validate the returned JSON with
`prd_reporting.py validate-result` before showing it; a delegated result never
writes project state.

Apply `reporting.delegation.fallback` when delegation is disabled, AI-Collab is
unavailable, times out, or returns an invalid result: `main` uses the current
host model for the summary, `deterministic_only` reports the generated facts
without model-written narrative, and `fail` reports that the optional summary
is unavailable. The deterministic report files remain authoritative in every
mode.