The agent writes one Markdown report per session, in this shape. Keep the
finding categories (BUG/CONCERN/OBSERVATION/PASS/FAIL/BLOCKED)
consistent so reports are skimmable and comparable across runs.
# Playtest Report: <session name or "Exploratory">
**Date:** <YYYY-MM-DD>
**Target:** <target name>
**Personality:** <bug-finder | feature-tester | feel-tester>
**Account:** <username>
**Goals file:** <filename or "none">
**Duration:** <~minutes>, <N> commands
## Summary
<2-4 sentence narrative of the session arc.>
## Goal Results
(Omit this section for exploratory runs with no goals file.)
- [x] <goal id>: <do> — PASS: <evidence>
- [ ] <goal id>: <do> — FAIL: <observed vs. expected>
- [ ] <goal id>: <do> — BLOCKED: <why>
## Findings
### BUG: <title>
<what you did, what happened, what should have happened (with repro steps)>
### CONCERN: <title>
<what seemed wrong and why>
### OBSERVATION: <title>
<notable behavior worth recording>
### PASS: <title>
<a feature/experience confirmed working>
## Stats
- Commands sent: <N>
- Errors seen: <N>
- Bugs / Concerns / Observations: <N> / <N> / <N>- One report per session; name it
<date>-<target>-<personality>[-<goals>].md. - Every BUG should be reproducible from the report alone: where, what you typed, what happened, what you expected.
- For goal runs, every goal in the goals file gets a line in Goal Results.
- Keep findings concrete; avoid vague "feels off" without saying why.