Agent Operator Skills

Reusable AI-agent skills for planning, audits, security reviews, changelogs, CLIs, SEO, and multi-step coding loops.

Why This Exists

Most agent prompts are too vague. They ask an AI to "be careful" or "do better" without giving it a repeatable operating pattern.

This repo is a curated set of public skills that make coding agents more disciplined:

• break large work into reviewable chunks
• force verification before claiming completion
• audit high-stakes output for weak assumptions
• generate release notes and changelogs from real evidence
• run technical SEO and security workflows with concrete checks
• keep private names, accounts, local paths, and customer data out of public skills

Star this repo if you want a clean reference set for building safer, more useful agent workflows.

Included Skills

Skill	Best For
codex-improvement-loop	Coordinating large coding initiatives with scoped workers, reviewers, loop state, and verification gates.
honest-output-auditor	Auditing important outputs before calling them done.
security-best-practices	Applying framework-specific secure-coding guidance for Python, JavaScript/TypeScript, Go, React, Next.js, Express, Django, Flask, FastAPI, Vue, and jQuery.
security-threat-model	Producing repo-grounded AppSec threat models with evidence anchors and abuse paths.
security-ownership-map	Finding ownership risk, bus factor, and sensitive-code stewardship gaps from git history.
seo-audit	Running and interpreting technical SEO audits for websites and local previews.
cli-creator	Building small, composable CLIs from APIs, scripts, docs, or admin workflows.
changelog-generator	Turning git history into user-facing changelogs and release notes.
define-goal	Turning fuzzy intentions into concrete, measurable agent goals.

Fast Start

Clone the repo:

git clone https://github.com/zgroves24/agent-operator-skills.git
cd agent-operator-skills

Copy one skill into your agent's local skill directory:

cp -R skills/honest-output-auditor ~/.codex/skills/

Or copy the whole public set:

cp -R skills/* ~/.codex/skills/

Other agent systems can use the same folders if they support a SKILL.md style layout.

Recommended Combos

Goal	Skills
Big codebase improvement	codex-improvement-loop plus honest-output-auditor
Security review	security-best-practices, security-threat-model, security-ownership-map
Release prep	changelog-generator plus honest-output-auditor
Website quality pass	seo-audit plus honest-output-auditor
Turn a repeated workflow into tooling	define-goal plus cli-creator

What Makes These Different

Each public skill is expected to have:

• a clear trigger
• concrete workflow steps
• verification gates
• blocked-state behavior
• no private account assumptions
• no personal data
• no product-specific claims

See docs/skill-quality-rubric.md for the bar used before publishing a skill here.

What Is Not Included

This repo intentionally excludes:

• personal admin workflows
• private product workflows
• outreach, analytics, ads, vault, or screen-history workflows tied to one local setup
• vendor-owned skills already published elsewhere
• private names, emails, paths, customer data, tokens, cookies, or secrets

See SANITIZATION_AUDIT.md and PROVENANCE.md for the publication notes.

Assets

• assets/agent-operator-skills-mark.svg is the lightweight repo mark.
• assets/agent-operator-skills-logo.png is the generated launch logo.
• assets/agent-operator-skills-profile.png is a square profile-size version.

Contributing

See CONTRIBUTING.md. New skills should be portable, specific, audited, and useful outside one private workspace.

License

Apache-2.0. See LICENSE.

This is not an official OpenAI, Anthropic, Vercel, GitHub, or platform-owned repository.