A catalog of agent artifacts — agents, skills, rules, scripts, and profiles
This repository also configures its own agents. See what agent-tools tells them →
Skill Claude CodeCodex
Run an accessibility audit — keyboard nav, ARIA, contrast, focus, labels, landmarks — and report gaps against WCAG 2.2 AA. Reports, does not auto-fix. Automation-first; asks only for the target surface. Use whenever the user wants an accessibility audit, a WCAG check, or runs /a11y, or after building any UI component.
Skill Claude CodeCodex
Think through what you are about to build like a senior engineer before writing any code. Surfaces decisions, aligns on language, and produces a clear implementation plan written to context/plan.md. Automation-first — answers its own questions from the context system and only surfaces genuinely blocking decisions. Use…
Skill Claude CodeCodex
Run a focused tech-debt cleanup pass — dead code, duplicate helpers, stale deps, unused exports, drift from conventions — and produce a safe, ordered removal plan. Automation-first; asks only before destructive removals. Use whenever the user wants a cleanup pass, to reduce tech debt, or runs /cleanup. Distinct from…
Skill Claude CodeCodex
Scaffold and continuously improve a project's shared context system — a context/ folder plus AGENTS.md, CLAUDE.md, and README.md — so AI agents build within a defined, accurate, modern structure instead of improvising. Self-improving; every run and every use makes the context more detailed and more true.…
Skill Claude CodeCodex
After building any UI component, extract the visual patterns that matter for consistency and save them to context/memory/ui-registry.md, so every component built afterward matches what came before. Uses context/design-system.md (tokens, rules) as the standard and context/design/ as the reference. Runs automatically…
Skill Claude CodeCodex
Write a postmortem from an incident timeline and contributing factors — timeline, impact, root cause, contributing factors, action items with owners. Automation-first; asks only when the timeline is unclear. Use whenever the user wants to write a postmortem, an incident report, or runs /postmortem, after /recover has…
Skill Claude CodeCodex
When something goes wrong during a build, diagnose what type of failure it is before deciding how to respond — targeted fix, hard reset, or full rethink. Uses the context system to tell a local bug apart from a polluted session or a wrong foundation, and routes a rethink back through /architect. Automation-first …
Skill Claude CodeCodex
Save what matters at the end of a session so the next one picks up exactly where you left off, and restore context at the start so nothing is lost. Stores state in context/memory/progress.md and reads the context system on restore. Automation-first — in automated flows it saves without prompting. Use whenever the user…
Skill Claude CodeCodex
After building a feature, verify it matches what was planned, respects the system architecture and design standards, and is ready for production. Checks the build against the context system (plan.md, architecture.md, standards.md, design-system.md, ui-registry.md) and reports issues with severity so the developer…
Skill Claude CodeCodex
Set up agents for the project the way you ask — tell it to set up subagents and it writes Claude Code .claude/agents files (or the host equivalent), tell it to create an agent team and it writes a Claude Code Agent Team spawn template, or do both — all derived from the project's systems and saved state. Also resumes…
Skill Claude CodeCodex
Pressure-test a project against the right quality bar and the right approach for what it is trying to be — not just against its own rules. Establishes the target tier, researches how the best comparable projects are built, reasons from first principles, and reports where the project or its context falls short or uses…
Skill Claude CodeCodex needs its repo
Use when the user asks to enhance, improve, rewrite, optimize, or clarify a prompt — or when always-on prompt enhancement is enabled for the session. Rewrites casual or vague chat into precise, portable agent instructions without injecting workspace paths, stack, or session-local context. Supports usage modes: auto…
Skill Claude CodeCodex needs its repo
Install enhance-prompt and related prompt skills, rules, and agents into a project (normal or auto/always-on). Use on /enhance-prompt-setup, enhance-prompt setup, or when wiring prompt tooling like agent-tools.
Skill Claude CodeCodex
Estimate the complexity and effort of a task before commitment. Sizes work into T-shirt sizes or hours, names the riskiest parts, and flags work that should be split. Automation-first; asks only when the task is ambiguous. Use whenever the user wants an estimate, a size, a risk check, or runs /estimate, or before…
Skill Claude CodeCodex
Set up a GitHub repository the way this catalog and agentry do — security policy, quality CI, Dependabot, branches, tag/branch rulesets, README/releases, and community files (issues, PRs, discussions). Use when creating or hardening a GitHub repo, or on /github-setup.
Skill Claude CodeCodex
Cut a release end-to-end — generate the changelog from commits, bump the version per the project's scheme, tag, and produce release notes. Automation-first; asks only for the bump type or target version. Use whenever the user wants to cut a release, ship a version, generate a changelog, or runs /release.
Skill Claude CodeCodex
Scaffold a new feature end-to-end in one pass — module, types, tests stub, docs stub, and registration — so the first commit isn't boilerplate drift. Automation-first; infers the project's conventions from neighbors. Use whenever the user wants to scaffold a feature, stub a module, or runs /scaffold, or before…
Skill Claude CodeCodex
Triage a list of bugs or issues by severity and root-cause guess before a human touches them. Assigns priority, names the likely failing system, flags duplicates and dependencies, and proposes a kill order. Automation-first; asks only when an issue is ambiguous. Use whenever the user wants to triage bugs/issues, sort…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: