wimpysworld/nix-config/review-report-path
Claude Code and agent skill by wimpysworld/nix-config
- community
Checking that the SKILL.md file is still in the repository.
Install
npx vskill@latest install wimpysworld/nix-config/review-report-pathSKILL.md
Review Report Path
Store private diagnostic reports from reviews, audits, and analysis in durable per-user state storage by default:
The calling command supplies <report-name>. This skill supplies <project>, <target>, and <run-id>.
This convention excludes project documents, tracker content, disposable plans, transport payloads, and runtime logs. Keep those outputs under their own workflow contracts.
Honour explicit user destinations unless the calling workflow restricts the input or output boundary. Otherwise, keep reports outside the repository and never commit them. Do not use /tmp or $TMPDIR for default report storage. A fenced process can have private temporary storage that disappears after that process exits.
Project
<project> is the repository directory name, kebab-case. In a git worktree, take the name of the main repository, not the worktree directory, so every review of one repository lands under one directory.
Target
<target> names what is under review. It keeps concurrent reviews apart, so two reviews running at once never write to the same file.
Take the first rule that matches:
| Target | Slug | Example | |
|---|---|---|---|
Pull request, by URL, number, or owner/repo#123 | pr-<number> | pr-123 | |
Branch or worktree whose branch has an open pull request, found with gh pr view --json number | pr-<number> | pr-123 | |
| Linear issue, by key or URL | issue-<key> | issue-eng-123 | |
GitHub issue, by URL or owner/repo#123 | issue-<number> | issue-456 | |
| Single commit | commit-<short-sha> | commit-a1b2c3d | |
| Branch with no pull request | branch-<name> | branch-feat-fix-auth | |
| Local file, such as a plan or pasted feedback | file-<basename> | file-plan | |
| Anything else, including a detached checkout | worktree-<dirname> | worktree-nix-config-pr-123 |
A command that takes no argument reviews the checkout it runs in, so it starts at the branch rules.
Normalise every slug: lowercase it, replace each character outside a-z0-9 with a hyphen, collapse repeated hyphens, trim leading and trailing hyphens, and cut it to 60 characters.
Run
Only the owner starting a new report-writing workflow allocates a run. Lookup creates no directories or files. A worker assigned a complete review is that owner and allocates one exclusive run when none is supplied. When the parent supplies a run, the complete-review worker reuses it. A review-lane worker always uses the parent's supplied run and fallback paths, never a new run.
For a new report-writing workflow, derive <project> and <target>, then allocate one run before any worker writes fallback findings:
mktemp -d creates the run directory exclusively. Keep the generated directory name as <run-id>. Use the same run_dir for the final report and all fallback findings from that invocation.
Find a Report
Derive the target directory from durable state, then search every run:
When the caller does not name a report, exclude worker fallback files:
The run directory stays in each result, so the caller can select an exact past run.
Rules
- Create missing state and target directories only when allocating a new report run.
- Never delete a run directory, report, or fallback findings file. Never reuse a run directory for a later invocation.
- Never overwrite an existing report or findings file. If an expected new path exists, stop writing to it. The workflow owner allocates another run and supplies replacement paths before workers write.
- Report the written path in your output so the user can find it.
- To find a report, derive the same project and target directory in per-user state storage. Search all
run-*directories under that target. - When the caller names a report file, list each matching
<run-id>/<report-name>. Use the sole match, or ask which run to use when several match. - When the caller does not name a report file, list the reports with their run IDs. Use the sole report, or ask which one to use when several exist.
- Unless the calling workflow explicitly overrides selection, never select the newest report when several reports match. Ask the caller to select an exact run.
- When no matching report exists, list available target directories under
${XDG_STATE_HOME:-${HOME}/.local/state}/agent-reviews/<project>/and stop instead of guessing. This includes missing, empty, and fallback-only target directories.
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckinghome-manager/_mixins/agentic/assistants/skills/review-report-path/SKILL.md
Popularity
- Repository stars
- 729
- Repository forks
- 33
- 7-day trend score
- 0
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/wimpysworld/nix-config/review-report-path)