playtest-prd-prep
Reconcile one validated Continuity Loom author-playtest report against live repo and tracker state, preserve proven strengths, and write one same-stem change portfolio for playtest-to-issues custody before any later /to-prd pass. Invoke explicitly with a reports/playtest-*.md path; this skill diagnoses and prepares but does not implement or publish.
- License 许可
- GPL-3.0
- 最近更新
- 2026-11-10
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- playtest-prd-prep
- description
- Reconcile one validated Continuity Loom author-playtest report against live repo and tracker state, preserve proven strengths, and write one same-stem change portfolio for playtest-to-issues custody before any later /to-prd pass. Invoke explicitly with a reports/playtest-*.md path; this skill diagnoses and prepares but does not implement or publish.
- disable-model-invocation
- true
02 资源文件 4
- 参考文档agents/openai.yaml285 B
- 参考文档references/prep-format.md11.9 KB
- 脚本scripts/validate-prd-prep.mjs48.8 KB
- 脚本scripts/validate-prd-prep.test.mjs35.2 KB
03 SKILL.md 正文
Playtest PRD Prep
Turn one .claude/skills/playtest report into
reports/<playtest-report-stem>-prd-prep.md. This run may write only that prep artifact. Never
edit the source report, product code, tests, docs, skills, specs, tickets, or tracker; never invoke
/to-prd or satisfy its seam checkpoint. The validated artifact goes through
$playtest-to-issues before any later /to-prd pass.
Intake
Require one canonical reports/playtest-<story-slug>-<timestamp>.md path or an unambiguous glob.
Exclude assets and derivative *-prd-prep.md files. If needed, search reports/
case-insensitively and proceed only when exactly one report resolves.
Run:
node .claude/skills/playtest/scripts/validate-report.mjs --report <report>
node .claude/skills/playtest-prd-prep/scripts/validate-prd-prep.mjs --inspect-source <report>
Stop without writing when schema, privacy, or ledger failures make exhaustive disposition unsafe.
Continue when the inspector explicitly classifies a bounded historical prompt-evaluation mismatch
as nonblocking; disclose it and add the required $skill-audit maintenance follow-up without
repairing the source. Missing optional evidence, stale commits, unavailable tracker reads, and
failed reproduction are limitations, not source invalidation; affected items become verification
work.
Before analysis, record the canonical path, validation and inventory, run mode, prior-report pointer and prior prep state, current same-stem prep state, branch, HEAD, and baseline worktree.
Reconcile
1. Establish live evidence
- Retain exact
git status --short --untracked-files=all,git branch --show-current, andgit rev-parse HEADoutput. Pre-existing dirt remains unowned unless this run explicitly adopts it. - Read
AGENTS.md,docs/ACTIVE-DOCS.md,docs/principles/FOUNDATIONS.md, affected active authorities,docs/agents/issue-tracker.md,docs/agents/triage-labels.md, andtickets/README.md. Usearchive/**only when explicitly routed as historical evidence. - Read the complete report and use its Cumulative Finding Ledger as the frontier. Follow only its
explicit
prior_reportchain, and only when anopen,repeated, ornot-retestedrow lacks decision-relevant detail. Never scan unrelated reports. - Compare the launch HEAD with the current affected paths. Distinguish relevant committed drift, relevant worktree drift, and unrelated dirt. Classify report durability with tracked, clean, publication-ref-visible, and content-identical evidence; summarize but never cite machine-local evidence.
- For a continuation, derive the prior report's same-stem prep. Treat it as historical
recommendations, not authority. If present, reconcile its first action, every PRD candidate,
and every non-PRD item as
consumed,still live,rejected, orsuperseded; otherwise recordmissing at intake. - Apply the same reconciliation to any current same-stem prep before reusing it.
- Search exact report terms and IDs across current open and relevant closed tracker work. Start with projected metadata; fetch bodies/comments only for likely owners. If unavailable, record that limitation and do not claim AFK-ready label certainty.
2. Disposition the report
Give every cumulative row exactly one disposition from the artifact contract; every prioritized ID must already occur there. Treat observations as evidence, not mandates. Promote a stable rule only when supported by current source/tests, repeated evidence, deterministic reproduction, or governing authority. Preserve one-run ambiguity and model variability as verification, coverage, or research unless a safe deterministic mitigation is justified.
Use evidence in this order: report and privacy-safe evidence; current authorities, implementation,
and focused tests; exact tracker overlap/closeout; focused read-only probes; then a browser
reproduction only if contradiction still blocks classification. For that last case, read
browser-driver.md, keep blank-key guarded localhost
state under /tmp, and do not start a full journey, accept prose, click provider controls, or make
OpenRouter requests.
Every strength is a preservation constraint, mapped globally or to the work it constrains with regression evidence. It never creates scope. Finish with one supported disposition, destination, and PRD impact for every row.
3. Build the smallest governed portfolio
Trace each report-derived rule across mechanically affected code, tests, active docs, and skills; exclude work that does not trace to a row, strength, or required consistency consequence. Route:
- broad or risky product behavior to a PRD candidate;
- narrow scoped product work to a ticket candidate;
- playtest/report process work to skill maintenance;
- inaccurate active authority to a doc correction;
- conflict with closed work to verification/reopen;
- insufficient evidence to coverage or research; and
- covered, resolved, preference-only, duplicate, or disproportional work to covered/rejected.
Bundle PRD rows only when product rule, decision, implementation seam, and acceptance proof align. Default to one first PRD plus deferred follow-ons; use a multi-PRD program only for real dependency or shared ratification. No new PRD is valid.
Rank work by invariant risk, author trust/core-loop impact, severity, dependency, evidence,
testability, and strength-regression risk. First operational action is the first substantive
action custody must resolve—not $playtest-to-issues, playtest-to-issues, or /to-prd; use
none - <reason> when nothing precedes the remaining portfolio.
Describe behavior, authority, current seam, surfaces, acceptance, and testing at PRD-ready depth. Avoid patch order, volatile symbols, components, or abstractions unless architecture makes them unavoidable. If one material scope, sequence, or stewardship choice remains, ask one focused question with a recommendation; do not run a routine requirements interview.
4. Write the artifact
Read the artifact contract in full immediately before writing. Write or rewrite the one default same-stem artifact at the contract's current version; never fork numbered/timestamped copies. The producer owns migration of older or unversioned prep artifacts and must account for all prior recommendations.
Consult /to-prd only for house style. Do not draft a PRD, create an issue, apply labels, or ask
its seam question; record that the checkpoint remains owed. Keep the artifact privacy-safe: no full
prompts, payloads, raw assistance, candidate/accepted prose, keys, localhost URLs, or machine-local
paths. Summarize temporary evidence as summarized, not cited.
The artifact must let /to-prd select and draft the intended package without rereading the source
for provenance, scope, strengths, authorities, sequencing, or deferrals.
5. Validate and close
Keep the contract's three self-checks at their draft values and run:
node .claude/skills/playtest-prd-prep/scripts/validate-prd-prep.mjs --draft <report> <prep-artifact>
Fix every error and compare emitted counts with the report. The helper proves shape/coverage, not semantics. Manually review dispositions, grouping, authorities, preservation, evidence, privacy, and stale language.
Immediately before the freshness check, capture full status and branch again, replace the final worktree ledger with that exact snapshot, and classify every path. Then set the three final self-check values and run:
node .claude/skills/playtest-prd-prep/scripts/validate-prd-prep.mjs <report> <prep-artifact>
After a pass, capture status and branch again. If either differs from the validated snapshot, update the ledger, repeat affected semantic/privacy/freshness review, and revalidate until equal. Do not run root gates for this report-only change; report focused checks and skipped broader gates.
Completion requires an accepted source inspection, exhaustive row and strength coverage, passing
final validation, completed semantic/privacy/freshness review, an exact worktree boundary, and the
$playtest-to-issues handoff.
Render this exact keyed block from the validated artifact as the final content:
Source report: <canonical source report>
Prep artifact: <same-stem prep artifact>
First operational action: <substantive action or none - reason>
Publication verdict: <package and intended first PRD, or no-new-PRD result>
Deferred work: <every deferred PRD candidate, or none>
Non-PRD work: <every item and destination, or none>
Issue-custody handoff: $playtest-to-issues "<prep-artifact>"; required before /to-prd
Tracker and durability: <tracker limitations; source durability; artifact durability>
Intentional change: <authored prep artifact>
Remaining dirt:
- <path>: <pre-existing, concurrent/unowned, or intentional classification>
Non-actions: no implementation, source-report edit, tracker mutation, PRD publication, or /to-prd seam checkpoint occurred
04 同仓库更多 Skills
triage
Move issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.
triage
Move issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.
triage
Move issues and external PRs through a state machine of triage roles — categorise, verify, grill if needed, and write agent-ready briefs.
to-prd
Turn the current conversation into a PRD and publish it to the project issue tracker — no requirements interview, just synthesis of what you've already discussed; the seam confirmation is the sole checkpoint.
to-prd
Turn the current conversation into a PRD and publish it to the project issue tracker — no requirements interview, just synthesis of what you've already discussed; the seam confirmation is the sole checkpoint.
to-prd
Turn the current conversation into a PRD and publish it to the project issue tracker — no requirements interview, just synthesis of what you've already discussed; the seam confirmation is the sole checkpoint.