agentleFS
jqaisystemsjqaisystems/jqai-ai-skills/skills/github-safe-publisher/SKILL.md

github-safe-publisher

Skill3 starsChanged 6 months agoMITView on GitHub

Prepare private work for safe public GitHub publishing. Use when the user asks to publish safely, turn private project work into a GitHub-safe artifact, sanitize a project, make a public case study, prepare material for GitHub, review a repo before pushing, or check whether files contain secrets, client data, raw prompts, logs, local paths, private source, or other sensitive material.

  • Reads credentials

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 7 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Very specific:

    Read references/public-safety-boundaries.md before deciding what can be reused, generalized, or removed.

    line 15
  2. Distinctive:

    Remove or replace sensitive details: credentials, account data, client material, raw prompts, logs, databases, exports, local paths, …

    line 17
  3. Distinctive:

    Run the scanner on the proposed public folder:

    line 18
  4. Very specific:

    Read references/release-workflow.md before staging changes.

    line 24
  5. Distinctive:

    If blocked, explain the minimum fix.

    line 35
  6. Very specific:

    Use scripts/scan_public_safety.py for a deterministic first pass.

    line 47
  7. Distinctive:

    Never run broad staging commands such as git add .

    line 65
  8. Distinctive:

    Prefer staging named files only after inspecting them.

    line 66
  9. Distinctive:

    Do not commit or push unless the user explicitly asks for that action after the safety review is clean.

    line 75
  10. Distinctive:

    When preparing or reviewing a public candidate, respond with:

    line 79

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

GitHub Safe Publisher

Overview

Use this skill to turn private project material into public-safe GitHub artifacts such as case studies, skill releases, README updates, demo notes, and example prompts. The skill produces a safety verdict and concrete next steps; it must not auto-commit, auto-push, publish, upload, send, or share anything without explicit human review.

Workflow

  1. Identify the intended public artifact: skill, case study, README section, demo screenshot note, example prompt, template, or changelog.
  2. Read references/public-safety-boundaries.md before deciding what can be reused, generalized, or removed.
  3. Work from a copy or rewritten summary, not the private original. Prefer a sanitized draft folder before touching a public repo.
  4. Remove or replace sensitive details: credentials, account data, client material, raw prompts, logs, databases, exports, local paths, private source, and screenshots with private UI.
  5. Run the scanner on the proposed public folder:
python scripts/scan_public_safety.py path/to/public-candidate --no-baseline
  1. Read references/release-workflow.md before staging changes.
  2. Require manual review of git status --short, the unstaged diff, and the staged diff before any commit or push.

Verdicts

Return one of these verdicts in every review:

  • BLOCK: Real secrets, credentials, raw private data, client-identifying material, local machine paths, databases, exports, logs, or unintended source files are present.
  • REVIEW: The artifact is probably publishable after specific human checks or rewrites.
  • READY: The artifact uses public-safe examples, passes the scanner, and has a clean manual diff review.

If blocked, explain the minimum fix. Do not rewrite blocked material into a public repo until the sensitive source is removed or generalized.

Sanitization Rules

  • Replace real people, businesses, accounts, emails, phone numbers, domains, addresses, and project codenames with neutral examples.
  • Replace credentials with placeholders such as YOUR_API_KEY_HERE.
  • Convert raw private notes into public explanations; do not copy memory files, chat logs, prompts, or operational strategy verbatim.
  • Publish patterns, workflows, and lessons. Keep implementation source, deployment details, private data, and system-of-record content private unless the user explicitly confirms they are public-safe.
  • For screenshots, crop or recreate them with fake data. Do not publish account menus, browser tabs, analytics, emails, invoices, API dashboards, or file paths.

Scanner

Use scripts/scan_public_safety.py for a deterministic first pass. Treat it as a guardrail, not proof of safety.

Recommended commands:

python scripts/scan_public_safety.py path/to/candidate
python scripts/scan_public_safety.py path/to/candidate --no-baseline --json
python scripts/scan_public_safety.py path/to/candidate --no-baseline

The scanner checks risky filenames, file extensions, binary-like artifacts, local path patterns, credential terms, token formats, private-key blocks, and private/client wording. Folder scans use .public-safety-baseline.json when present to suppress known review findings; use --no-baseline to see the raw scan. A clean scan still requires human diff review.

Git Safety

Scanner reports withhold matched text and absolute input locations. Relative filenames may still identify private work, so keep reports private. Linked inputs are refused; unreadable content and incomplete enumeration prevent readiness. Excluded build/dependency folders and binary content are not a complete content audit.

The baseline file itself is scanned. Existing review fingerprints remain compatible for unchanged content, but blocking findings are never suppressed. Inspect matches privately at their locations rather than copying sensitive lines into chat or logs. Review the exact proposed release file list separately.

  • Never run broad staging commands such as git add . unless the user has already reviewed the whole repo state.
  • Prefer staging named files only after inspecting them.
  • Always inspect:
git status --short
git diff -- path/to/file
git diff --cached
  • Do not commit or push unless the user explicitly asks for that action after the safety review is clean.

Output Format

When preparing or reviewing a public candidate, respond with:

  1. Verdict: BLOCK, REVIEW, or READY.
  2. Public artifact type and target repo/folder.
  3. Files inspected or created.
  4. Scanner result summary.
  5. Remaining manual checks.
  6. Exact next commands, if any, without auto-running publish actions.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.