ワンクリックで
report-findings
Review and submit anonymized harness findings as GitHub issues to the forge repo. Opt-in, user-confirmed, no secrets or PII.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Review and submit anonymized harness findings as GitHub issues to the forge repo. Opt-in, user-confirmed, no secrets or PII.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Switch model mid-session preserving thinking blocks. Used by /model command and by BRD §6.2 failover when the primary provider rate-limits or fails.
Six-phase ReAct loop (pre-check, thinking, self-critique, action, tool, post). Activates per-workflow via the thinking_level knob in config/workflows.yaml. Replaces standard ReAct for workflows where independent verification of the reasoning improves outcomes.
Mines completed sessions for repeating {tool sequence → outcome} tuples. Scores by frequency × success_rate × novelty. High-scoring tuples become "instincts" — candidate skill seeds — in instincts/pending/.
Walk-back algorithm for the Spec-Auditor subagent. Given a failure at phase N, identifies the earliest phase whose spec, if tightened, would have prevented the failure. Proposes a surgical amendment.
Sessions stored as trees (not lists). /fork creates a branch from any point; /tree navigates; /branch labels a path; /export produces HTML for review. All branches live in one session file under sessions/<project>/<session_id>.json.
Progressive context refinement for subagents. Don't dump everything; retrieve in passes. Start with a list of file paths and one-line abstracts; load full content only when the abstract proves insufficient.
| name | report-findings |
| description | Review and submit anonymized harness findings as GitHub issues to the forge repo. Opt-in, user-confirmed, no secrets or PII. |
| argument-hint | [--enable | --disable | --dry-run] |
Submit anonymized harness findings from your project back to the forge repo as GitHub issues. This is opt-in only and always requires user confirmation before anything leaves the machine.
| Data | Example |
|---|---|
| Gate name + pass/fail | evaluator: FAIL |
| Error category | console-error, type-error |
| Pattern (truncated 200 chars) | Missing CSRF token on /api/... |
| Iterations to fix | 3 |
| Stack type | nextjs-postgres |
| Forge version | 2.0.0 |
| Hook violation type | secret-detected, file-too-long |
| Learned rule text | Always add CORS headers for... |
| Data | Reason |
|---|---|
| Source code | Never collected |
| File paths | Stripped during sanitization |
| API keys, tokens | Regex-stripped (sk-, ghp_, base64) |
| Emails, IPs, names | PII scan patterns from pii-scan.js |
| Repo name, org name | Not included in findings schema |
| Environment variables | Never collected |
| .env contents | Never collected |
Read project-manifest.json and check findings_reporting.enabled.
If the user passed --enable: set findings_reporting.enabled: true in the manifest and confirm. Exit.
If the user passed --disable: set findings_reporting.enabled: false in the manifest and confirm. Exit.
If findings_reporting.enabled is not true, print:
Findings reporting is not enabled. Run
/report-findings --enableto opt in.
Exit.
Read .claude/state/harness-findings-log.json. If the file does not exist or is empty, print:
No findings recorded yet. Findings are collected automatically during gate runs when reporting is enabled.
Exit.
Filter the array to entries where reported: false. If none remain, print:
All findings have already been reported.
Exit.
Apply the same sanitization patterns used by detect-secrets.js and pii-scan.js:
/sk-[a-zA-Z0-9]{20,}/g (OpenAI keys)/ghp_[a-zA-Z0-9]{36,}/g (GitHub PATs)/[A-Za-z0-9+\/]{40,}={0,2}/g (long base64 blobs)/[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}/g/\b\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}\b/g/\/[^\s:]+\.(ts|js|py|go|rs|java|json|yaml|yml|toml)/gpattern field to 200 charactersIf sanitization removes all meaningful content from an entry, drop it.
Write harness-findings.md in the project root using the template from .claude/templates/harness-findings.template.md.
Populate the following sections:
Display the full content of harness-findings.md to the user.
--dry-run was passed: print "Dry run complete. No issue created." and exit.Run:
gh issue create \
--repo {target_repo} \
--title "Harness Findings: {stack_type} {date}" \
--body-file harness-findings.md \
--label "harness-finding"
Where {target_repo} is read from findings_reporting.target_repo in the manifest (defaults to the forge repo).
If gh is not installed or not authenticated, print:
GitHub CLI is not available or not authenticated. You can install it at https://cli.github.com/ and run
gh auth login. The report has been saved toharness-findings.md— you can submit it manually.
Do not fail the skill.
findings_reporting.last_reported in the manifest to the current ISO timestampharness-findings-log.json as reported: trueharness-findings.md (the staging file)--dry-run flag is useful for reviewing what would be submitted without creating an issue.reported: true with a note.harness-finding label must exist on the target repo. If it does not, gh will prompt to create it — accept the prompt.category + pattern are deduplicated in the report (counted, not repeated).