GitHub-ready distribution package
Agent Skill Safety Kit, packaged like a repo.
A forkable/shareable repo bundle for reviewing AI-agent skills before install, publication, marketplace listing, or delegation. It covers tool authority, memory policy, receipts, rollback, evals, exceptions, and portability.
ZIP size: 46,454 bytes · SHA256: 1e284baaa3db77f8760f359dd633ef81264a055bb650a484ad25c7df32fdac15
What is inside
Review kit
30-minute checklist, permission matrix, verdict template, marketplace checklist, and portable skill contract.
30-minute checklist, permission matrix, verdict template, marketplace checklist, and portable skill contract.
Repo scaffolding
README, CONTRIBUTING, SECURITY, PR template, issue template, submission packet, .gitignore, and validation workflow.
README, CONTRIBUTING, SECURITY, PR template, issue template, submission packet, .gitignore, and validation workflow.
Distribution assets
Directory/community copy, visual preview pack, SVG cards, sample mini-report, redaction-friendly intake, vendor questionnaire, and GitHub publish packet.
Directory/community copy, visual preview pack, SVG cards, sample mini-report, redaction-friendly intake, vendor questionnaire, and GitHub publish packet.
Permission manifest schema + validator
Machine-readable permission schema, example JSON, template, and browser-local validator link for registry/marketplace review.
Machine-readable permission schema, example JSON, template, and browser-local validator link for registry/marketplace review.
Suggested publish flow
unzip agent-skill-safety-github-repo.zip
cd agent-skill-safety-kit
git init
git add .
git commit -m "Initial Agent Skill Safety Kit"
# create a public repo, then:
git remote add origin git@github.com:<owner>/agent-skill-safety-kit.git
git push -u origin main
Safety rule: do not paste secrets, private prompts, credentials, customer data, or proprietary internals into public repo issues, examples, or review cases. Use fictional or redacted examples only.
Suggested repo metadata
- Name:
agent-skill-safety-kit - Description: Dependency-style safety review kit for AI-agent skills: permissions, memory, receipts, rollback, evals, and marketplace readiness.
- Topics:
ai-agents,agent-safety,agent-skills,claude-code,openclaw,ai-security,tool-permissions,agent-memory,evals
When to get the $29 review
Run the free repo kit first. If the skill will touch real user data, browser sessions, SaaS systems, customer messages, durable memory, marketplace listings, money-adjacent records, or external actions, use the $29 Custom Agent Audit Mini-Report for an outside launch verdict. Start with the redaction-friendly intake; do not send secrets.