ai-plugins

A multi-harness marketplace of AI coding-assistant plugins for Claude Code, Codex, and other harnesses that adopt plugin or marketplace concepts.

Plugin catalog

Most plugins ship both a .claude-plugin/ and a .codex-plugin/ manifest and are registered in both marketplace manifests. Codex-only plugins are registered only in .agents/plugins/marketplace.json and intentionally omitted from the Claude Code marketplace. Provider-backed promptfoo evals exercise the relevant marketplace surface for each harness.

Claude Code

Plugin	Description	Version
worktrees	Goal-driven worktree setup plus a guard that blocks commits from the main checkout.	0.1.0
babysit-pr	Forge-agnostic PR/MR babysitting across GitHub, Forgejo, and GitLab.	0.1.0
engineering-standards	A stack-agnostic, portfolio-grade engineering regime: a guardrail skill and a scaffold skill.	0.2.0
agentic-systems-engineering	Portable guardrails for building, evaluating, and delivering LLM and agentic systems.	0.1.1
eval-case-reporter	Capture sanitized eval cases from bad or borderline AI-assistant behavior and submit them to this marketplace.	0.1.0

Codex

Plugin	Description	Version
worktrees	Goal-driven worktree setup plus a guard that blocks commits from the main checkout.	0.1.0
babysit-pr	Forge-agnostic PR/MR babysitting across GitHub, Forgejo, and GitLab.	0.1.0
engineering-standards	A stack-agnostic, portfolio-grade engineering regime: a guardrail skill and a scaffold skill.	0.2.0
agentic-systems-engineering	Portable guardrails for building, evaluating, and delivering LLM and agentic systems.	0.1.1
eval-case-reporter	Capture sanitized eval cases from bad or borderline AI-assistant behavior and submit them to this marketplace.	0.1.0
advisor	Read-only planning advisor for fuzzy tradeoffs, scope shaping, specs, and ticket plans.	0.1.0

When a plugin is added under plugins/, add catalog rows only for the harness marketplaces that list it. Codex-only plugins must not be added to the Claude Code table.

Using the marketplace (Claude Code)

Add this repository as a marketplace, then install a plugin from it:

# From inside Claude Code:
/plugin marketplace add jwilger/ai-plugins      # GitHub owner/repo shorthand
# ...or a local checkout:
/plugin marketplace add ./ai-plugins

/plugin install <plugin-name>@ai-plugins

The marketplace is referenced by its name (ai-plugins) in install commands, regardless of the URL you added it from. List and manage with /plugin list, /plugin marketplace update ai-plugins, and /plugin marketplace remove ai-plugins.

Using the marketplace (Codex)

Codex-facing marketplace metadata lives in .agents/plugins/marketplace.json, and each plugin has a .codex-plugin/plugin.json manifest. In a local checkout, install or sync the plugin from the matching directory under plugins/ using the Codex plugin flow available in your Codex environment.

The plugin names are the directory names, for example:

agentic-systems-engineering
eval-case-reporter
engineering-standards
babysit-pr
worktrees
advisor

Useful Codex entry points:

• agentic-systems-engineering: route LLM, agent, RAG, tool-use, structured output, provider-routing, observability, and delivery work through portable agentic-system guardrails.
• eval-case-reporter: recognize reusable failures or surprising assistant behavior, scrub sensitive data, preview the complete GitHub issue payload, ask before posting, and submit the sanitized issue with gh issue create.
• engineering-standards: apply the repository's broader engineering regime, including the no-force-push rule.
• advisor: delegate fuzzy planning, tradeoff analysis, scope/spec shaping, and ticket planning to a read-only advisor subagent.

Developing in this repo

A Nix flake provides a reproducible devshell with Node, npm, jq, prettier, ripgrep, fd, just, and bats.

nix develop        # enter the devshell
# or, with direnv:
echo "use flake" > .envrc && direnv allow

Any globally installed npm tooling (npm install -g …) is redirected into a git-ignored ./.dependencies/ directory by the devshell, so it never pollutes your home directory. Delete that directory any time for a clean slate.

This repo also has a committed package.json/package-lock.json for the local Promptfoo eval runner. node_modules/ is ignored and restored with npm ci; the eval scripts run that automatically when the Promptfoo, Codex SDK, or Claude Agent SDK packages are missing.

See AGENTS.md for how to author, validate, and publish a plugin.

Eval reports

The repo-owned eval dashboard is generated under site/evals/ by node scripts/evals/build-site.mjs. It is a local/static artifact for review and workflow uploads; the durable record is repo-owned and does not depend on promptfoo-hosted sharing.

Provider-backed Claude Code and Codex results require both repository secrets: OPENAI_API_KEY and ANTHROPIC_API_KEY. If either secret is absent, the live eval workflow skips provider-backed behavior evals instead of pretending they ran.

The dashboard includes latest-run status, provider/case/sample pass rates, threshold status, and plugin/skill summaries so regressions can be traced back to the marketplace surface they exercise.

The canonical promptfoo behavior evals run through Promptfoo's native coding agent providers: anthropic:claude-agent-sdk for Claude Code and openai:codex-sdk for Codex. The runner generates the promptfoo config from the current marketplace manifests, so each provider loads the full ai-plugins marketplace, not a single plugin in isolation. Routing and plugin composition are therefore part of the measured behavior. Promptfoo is pinned at 0.121.17; the Promptfoo, Codex SDK, and Claude Agent SDK packages are pinned in package.json and package-lock.json. The runner disables prompt response caching and hosted sharing so a behavior run is a fresh local record.

Default eval harness posture:

• Claude Code: anthropic:claude-agent-sdk, Sonnet 5 via the sonnet alias, local Claude Code authentication via apiKeyRequired: false, and all local plugins with skills: all. The intended human-facing Claude Code posture remains Sonnet high effort with Opus 4.8 advisor where that harness exposes those controls; Promptfoo's current Claude Agent SDK provider does not expose those knobs in this repo's generated config.
• Codex: openai:codex-sdk, gpt-5.5 with model_reasoning_effort=medium, read-only sandbox, no approvals, streaming, deep tracing, and a generated CODEX_EVAL_HOME containing every repo plugin. Model-graded assertions also use openai:codex-sdk by default so OpenAI model access goes through local Codex auth rather than OPENAI_API_KEY.

The canary suite is separate from behavior evals. Canaries may explicitly ask the harness to prove plugin and skill loading. Behavior prompts stay natural and do not tell the model to use this repository's plugins.

Repeated samples are a deliberate measurement choice, not a blanket rule. Use more distinct cases when estimating population quality; use repeated samples when measuring per-input reliability, pass@k capability, pass^k reliability, or small stochastic differences. Trusted release evidence for this repository defaults to EVAL_SAMPLES=3; PR dry-runs do not run live samples.

Pull-request CI validates the eval configuration with --dry-run but does not claim behavior evidence. Provider-backed behavior evidence comes from local, scheduled, manual, or main runs where Claude Code and Codex authentication are available.

To produce the same artifacts locally:

just evals  # runs provider-backed evals, shares the result, and prints the URL
nix develop -c scripts/evals/run.sh
nix develop -c scripts/evals/run.sh --suite canary
nix develop -c node scripts/evals/build-site.mjs

just evals uploads the latest eval result through promptfoo share. For a local-only report, run scripts/evals/run.sh and then nix develop -c node_modules/.bin/promptfoo view. If a behavior eval exits with Promptfoo's normal failure status after writing artifacts, just evals still attempts to share the report and then returns the original eval status. If the eval run is interrupted with Ctrl-C, just evals stops without sharing.

Codex users who install agentic-systems-engineering also get an optional Promptfoo MCP server (promptfoo@0.121.17 mcp --transport stdio). Use it for agent-assisted config validation, focused eval runs, result inspection, and fixture development. It supplements the canonical runner; it does not replace the repo-owned artifact path above. Promptfoo's separate mcp provider is for testing MCP servers as systems under test and should be added only when a plugin or project exposes an MCP server to evaluate.

Reporting eval cases

When a plugin, skill, prompt, or workflow behaves incorrectly or only partially works, file an Eval case issue in this repository. Eval cases are the intake path for future regression fixtures in evals/fixtures/.

Include the sanitized input, actual behavior, expected behavior, expected eval outcome (pass, fail, partial, adversarial, or unsure), and the assertion or rubric that would catch the behavior. Do not include secrets, credentials, auth headers, cookies, session ids, private keys, private client data, private repository names, internal hostnames, or raw proprietary source excerpts.

Repository layout

.
├── .agents/
│   └── plugins/
│       └── marketplace.json  # Codex-facing marketplace manifest
├── .claude-plugin/
│   └── marketplace.json      # Claude Code marketplace manifest
├── .github/
│   ├── ISSUE_TEMPLATE/       # eval-case intake form
│   └── workflows/            # CI and eval workflows
├── docs/
│   └── superpowers/plans/    # implementation plans for larger changes
├── evals/
│   ├── fixtures/             # behavior eval scenarios
│   └── promptfoo/            # promptfoo loaders and assertions
├── plugins/                  # one subdirectory per plugin
├── scripts/
│   ├── evals/                # eval config generator, runner, and dashboard builder
│   └── tests/                # Bats tests
├── site/
│   └── evals/                # generated dashboard target, ignored except .gitkeep
├── flake.nix                 # Nix devshell
├── AGENTS.md                 # guidance for AI agents working in this repo
└── README.md                 # this file

License

See individual plugins for their licenses.