원클릭으로 Manus에서 모든 스킬 실행

self-report

스타0

포크0

업데이트2026년 5월 9일 09:41

File GitHub issues against joestump/claude-skills when another skill in this repo hits friction during a run. Other skills in this repo delegate their self-evaluation reporting to this skill rather than implementing their own filing logic. Triggers when another skill explicitly invokes self-report after detecting that its trip thresholds were exceeded (tool-call budget, repeated bug retries, material assumptions, render integrity issues, user pushback in same turn), or when the user explicitly invokes /self-report. If multiple skills from this repo ran in the same session and each independently tripped, file one issue per skill — never bundle reports for different skills into a single issue. Each issue is labeled with both skill-self-report and skill:<skill-name> so they can be filtered per-skill on the issues tab. Issue bodies must never contain user data, account balances, names, or document contents — only meta-signals about how the run went.

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

joestump

joestump/claude-skills

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

self-report

This skill is the single owner of "skill hit friction → file a GitHub issue." Other skills in this repo do not implement their own filing logic. They detect their own trip conditions, then hand off to self-report with a skill name, a short title, and a body.

When to run

Another skill in this repo signals that its run tripped one or more thresholds defined in that skill's own references/self-eval.md. Common signals (each skill defines its own exact thresholds):
- Tool-call budget exceeded for the run mode
- Same bug retried more than twice
- More than 3 material assumptions during the run
- Render-integrity failures (literal ${...}, NaN/undefined in output, missing tabs, etc.)
- User pushback in the same turn ("wrong", "ugly", "broken", "fix this")
The user explicitly invokes /self-report.

Do not self-report this skill. If self-report itself fails, surface the failure to the user in chat. Filing-about-filing is a recursion trap.

Multi-skill runs

If two or more skills from this repo ran in the same session and each independently tripped, file one issue per skill. Never bundle reports for different skills into a single issue — they go to different audiences (the maintainer of each skill) and have different fix surfaces.

When multiple issues are filed in one session, mention this to the user once at the end:

I noticed friction in this run across retirement-plan and weekly-review. I've filed one issue per skill. [link, link]

Don't repeat. Don't nag.

Workflow

For each skill that tripped:

Build the issue body following references/issue-format.md. Lift the run-specific facts (tool-call count, retry count, phase reached, failure-signal checkboxes) from the calling skill's run state.
Sanitize. Apply references/privacy.md before writing the body — no balances, names, document contents, addresses, partner identifiers, or direct quotes from user docs.
File via the path that's available (in order):
1. GitHub MCP, if available. Repo: joestump/claude-skills. Labels: skill-self-report and skill:<skill-name>. Title: [<skill-name>] <short summary>.
2. gh CLI via scripts/file_issue.sh — call it once per skill: file_issue.sh <skill-name> "<title>" <body-file>. The script ensures the per-skill label exists, then files.
3. Fallback: the script writes the body to $SKILL_ISSUE_OUT_DIR/skill-issue-<skill-name>.md (default /mnt/user-data/outputs/) and prints a prefilled https://github.com/joestump/claude-skills/issues/new URL with labels=skill-self-report,skill:<skill-name>&title=... query params.
Mention to the user once per session, only on the fallback path:

I noticed friction in this run. I've written self-reports to skill-issue-*.md — paste them as new issues at https://github.com/joestump/claude-skills/issues/new if you'd like to track them.

Labels

skill-self-report — applied to every issue this skill files. Already exists on the repo.
skill:<skill-name> — per-skill filter. The script auto-creates this label on first use with color #c5def5. Examples: skill:retirement-plan, skill:weekly-review.

Things to never do

Never include user financial data, account balances, names, document contents, addresses, partner names, or direct quotes from user documents in an issue body.
Never bundle reports for multiple skills into a single issue.
Never file an issue for a clean run.
Never bring up the self-report more than once per session to the user.
Never self-report this skill.

이 저장소의 다른 Skills

같은 저장소

cut-release

joestump/claude-skills

Cut a release on GitHub or Gitea — the forge-level Release object (notes + assets) that sits on top of a git tag, NOT just the tag itself. Use this skill whenever the user says "cut a release", "cut a tag", "tag a release", "ship it", "publish v1.2.0", "do a release", or "draft release notes", and especially right after a PR is merged to the default branch when it's time to ship. The skill auto-discovers the forge from the git remote (github.com → GitHub, anything else → probe for Gitea/Forgejo), verifies CI is green on the merge commit before publishing, figures out the next semver tag from the repo's existing tag convention and the commits since the last tag, then creates an annotated tag and a Release with notes. Works for single-version repos and monorepos with per-component tags (like this one: `<skill-name>-vX.Y.Z`). Reach for this even if the user only says "tag" — confirm whether they want a bare tag or a full release.

2026-06-300

outline-runbook-curator

joestump/claude-skills

Curate, consolidate, and restyle the StumpCloud Runbooks wiki in Outline (outline.stump.rocks) against the repo as source of truth. Use this skill whenever the user wants to review/clean up the wiki, consolidate or reorganize runbooks, fix stale or duplicated docs, reconcile Outline against the repo (ADRs, inventory, docs/), restructure the doc tree into hierarchy-native nodes, or apply the domain color + topical icon scheme to wiki nodes. Trigger on phrases like "audit the runbooks", "clean up the wiki", "consolidate the Outline docs", "the wiki is out of date", "reorganize/restyle the Runbooks collection", or "color and icon the docs". This is the governance/curation counterpart to the `stumpcloud-omg` skill (which files individual incidents); use that for postmortems, this for whole-collection curation. Always propose a consolidation plan and get approval before any destructive merge/move/delete.

2026-06-270

refresh-miatrix-token

joestump/claude-skills

Refresh the Miatrix indexer API token in Prowlarr by logging into Miatrix via browser automation and extracting the current API key. Use this skill whenever Miatrix keeps re-downloading the same episodes or shows, when Prowlarr reports Miatrix indexer errors or failures, or when a Miatrix API key rotation is suspected. The skill handles the full workflow automatically: gathers credentials, logs into Miatrix, grabs the API key from the profile page, and updates Prowlarr's indexer config — no manual steps required.

2026-05-190

gemini-mockup

joestump/claude-skills

Generate high-fidelity UI mockup PNGs for any project by calling Gemini's image model through LiteLLM. Use this whenever the user asks to mock up, generate, regenerate, or update a UI screen, page, modal, or component — login screens, dashboards, settings panels, wizards, admin views, anything with a visual surface. The skill reads the project's visual identity from its `CLAUDE.md` at runtime, frames every output in macOS Safari chrome with a project-appropriate URL, saves to a stable filename so reruns overwrite cleanly, and scans in arrears for older mockups in the same directory that may be stylistically out of date. Prefer this over the bundled `mockup-generator` skill when LiteLLM is the gateway (env vars `OPENAI_BASE_URL` and `OPENAI_API_KEY` live in the user's login shell) and when project visual identity should drive style rather than light-mode DaisyUI defaults.

2026-05-130

retirement-plan

joestump/claude-skills

Generate a high-fidelity, durable retirement-plan.html from a Claude Project's financial documents and lifestyle assumptions. Triggers when the project name suggests retirement focus (retire, retirement, FIRE, early retire, freedom), the user invokes /retirement-plan or asks to build/update/refresh the plan, new financial documents are uploaded, the user mentions a money event (inheritance, RSU vest, property sale, large bonus, paid-off mortgage), or an existing retirement-plan.html is older than 3 months OR built from documents older than 6 months. Produces a single self-contained HTML file with five tabs (Execution plan, Income reality, Big purchases, At retirement, Lifetime projection), inline SVG chart, and interactive knobs. Re-triggers read the prior artifact as canonical state and apply deltas rather than starting over.

2026-05-090

name

self-report

description

self-report

When to run

Another skill in this repo signals that its run tripped one or more thresholds defined in that skill's own references/self-eval.md. Common signals (each skill defines its own exact thresholds):
- Tool-call budget exceeded for the run mode
- Same bug retried more than twice
- More than 3 material assumptions during the run
- Render-integrity failures (literal ${...}, NaN/undefined in output, missing tabs, etc.)
- User pushback in the same turn ("wrong", "ugly", "broken", "fix this")
The user explicitly invokes /self-report.

Do not self-report this skill. If self-report itself fails, surface the failure to the user in chat. Filing-about-filing is a recursion trap.

Multi-skill runs

When multiple issues are filed in one session, mention this to the user once at the end:

I noticed friction in this run across retirement-plan and weekly-review. I've filed one issue per skill. [link, link]

Don't repeat. Don't nag.

Workflow

For each skill that tripped:

Build the issue body following references/issue-format.md. Lift the run-specific facts (tool-call count, retry count, phase reached, failure-signal checkboxes) from the calling skill's run state.
Sanitize. Apply references/privacy.md before writing the body — no balances, names, document contents, addresses, partner identifiers, or direct quotes from user docs.
File via the path that's available (in order):
1. GitHub MCP, if available. Repo: joestump/claude-skills. Labels: skill-self-report and skill:<skill-name>. Title: [<skill-name>] <short summary>.
2. gh CLI via scripts/file_issue.sh — call it once per skill: file_issue.sh <skill-name> "<title>" <body-file>. The script ensures the per-skill label exists, then files.
3. Fallback: the script writes the body to $SKILL_ISSUE_OUT_DIR/skill-issue-<skill-name>.md (default /mnt/user-data/outputs/) and prints a prefilled https://github.com/joestump/claude-skills/issues/new URL with labels=skill-self-report,skill:<skill-name>&title=... query params.
Mention to the user once per session, only on the fallback path:

I noticed friction in this run. I've written self-reports to skill-issue-*.md — paste them as new issues at https://github.com/joestump/claude-skills/issues/new if you'd like to track them.

Labels

skill-self-report — applied to every issue this skill files. Already exists on the repo.
skill:<skill-name> — per-skill filter. The script auto-creates this label on first use with color #c5def5. Examples: skill:retirement-plan, skill:weekly-review.

Things to never do

Never include user financial data, account balances, names, document contents, addresses, partner names, or direct quotes from user documents in an issue body.
Never bundle reports for multiple skills into a single issue.
Never file an issue for a clean run.
Never bring up the self-report more than once per session to the user.
Never self-report this skill.