QuantEcon Skills - QuantEcon Skills

A Claude Code plugin marketplace housing QuantEcon’s shared agent skills and their supporting scripts.

Each plugin bundles one area of work — a skill (the instructions Claude follows) plus the deterministic scripts it drives — so the same versioned toolkit works locally for authors and RAs, and headlessly in CI.

📖 quantecon.github.io/skills — the documentation, rendered and navigable. It is built from the files in this repository, so reading either one gets you the same content.

Plugins¶

Plugin	For	Covers
`qe`	Authors and RAs writing lectures	Style checks against the QuantEcon style guide, and working through the review feedback on a PR once it is open
`benchmark`	Maintainers reviewing accelerated implementations	Measured, rubric-scored evaluation of a conversion
`audit`	Maintainers sweeping a whole repository	Bulk, read-only audits — every issue, every PR, a codebase, a translated series — each producing a written report

qe is the author-facing surface — one memorable prefix for everyday work, spanning a lecture’s life from drafting to merge. check-style is the umbrella (whole lecture, optional category filter, e.g. /qe:check-style lectures/aiyagari.md figures math) and the per-category sub-skills run the same shared rules individually; /qe:copilot-review picks the same lecture up after the PR is open, working through Copilot’s review comment by comment. benchmark and audit are specialist toolkits, installed by the maintainers who need them.

Which skills work right now is in Skill catalog — it lists what has merged and is operational, so this page does not repeat it. Skills whose scaffolding has merged but which do not run yet are in Using QuantEcon skills, because they still appear in the slash menu. Ideas nobody has committed to are in Future skill ideas.

Documentation¶

Guide	For
AGENTS.md	AI agents and contributors: canonical repo instructions — the single-source-of-truth principle, doc map, working conventions
Using QuantEcon skills	Authors/reviewers: setup, invoking skills, what to expect
Developing skills	Contributors: layout, conventions, dev loop, testing locally, versioning and releases, PR flow
Tutorial: run a full evaluation by hand	Tutorial: the evaluation procedure by hand, with the ge_arrow validation run as the checkable example
Tutorial: run a whole-tracker audit	Tutorial: a whole-tracker audit end to end, and how to review what it produces
benchmark plugin	The evaluation skill: review mode, triage mode, report format, manual pipeline
audit	The audit family: what belongs in it, the shared method, running one

Installation¶

Automatic (lecture repos)¶

Lecture repositories opt in by checking the following into their .claude/settings.json. Anyone who opens the repo and trusts it gets the marketplace and plugins installed automatically — no commands to run:

{
  "extraKnownMarketplaces": {
    "quantecon": {
      "source": { "source": "github", "repo": "QuantEcon/skills" }
    }
  },
  "enabledPlugins": {
    "qe@quantecon": true,
    "benchmark@quantecon": true
  }
}

audit is deliberately absent from the lecture-repo block: it is maintainer tooling, and the plugin is the enable unit, so auto-installing it would put org-wide audit skills in every author’s command list. Maintainers install it themselves.

Manual (any project)¶

/plugin marketplace add QuantEcon/skills
/plugin install qe@quantecon
/plugin install benchmark@quantecon
/plugin install audit@quantecon

Those are slash commands in a Claude Code session; restart the session afterwards, since plugins register at startup. /plugin is a terminal-CLI built-in, so in the VS Code extension or the web app use the equivalent claude plugin … CLI commands instead — see using-skills § Setup.

CI (GitHub Actions)¶

The official action accepts the marketplace and plugin directly:

- uses: anthropics/claude-code-action@v1
  with:
    plugin_marketplaces: "https://github.com/QuantEcon/skills.git"
    plugins: "benchmark@quantecon"
    prompt: "/benchmark:review-acceleration <args>"

Layout¶

.claude-plugin/marketplace.json   # the marketplace catalogue
scripts/                          # manifest + frontmatter validation, version-bump guard (run in CI)
qe/                               # author-facing plugin
  .claude-plugin/plugin.json
  CHANGELOG.md                    # what shipped in each version
  skills/check-style/SKILL.md     # umbrella skill
  skills/check-<category>/        # thin per-category sub-skills
  skills/copilot-review/SKILL.md  # PR review-feedback workflow
  references/rules/               # shared rule files (style-guide schema)
  scripts/                        # deterministic preflight checkers + fetch-copilot.sh
benchmark/                        # specialist plugin
  .claude-plugin/plugin.json
  CHANGELOG.md
  skills/review-acceleration/SKILL.md
  scripts/                        # supporting Python scripts the skill drives
audit/                            # maintainer-facing bulk-audit plugin
  .claude-plugin/plugin.json
  CHANGELOG.md
  skills/issues/SKILL.md          # one skill per audit subject
  references/                     # shared doctrine, org context, reporting guidance
  scripts/                        # deterministic snapshot + coverage machinery

Contributing¶

Open a PR adding or modifying a plugin directory and registering it in .claude-plugin/marketplace.json. Test locally with claude --plugin-dir ./<plugin> before submitting.

Run python scripts/validate.py before pushing — it checks that every plugin resolves, that plugin.json agrees with marketplace.json on name, version, and description, and that each SKILL.md has frontmatter whose name matches its directory. CI runs the same script on every PR; a malformed manifest otherwise breaks installation silently in every consuming lecture repository.

CI also fails a PR that changes files under a plugin directory without bumping that plugin’s version and adding its changelog entry — the install cache is keyed by version string, so an unbumped change reaches nobody (developing-skills § Versioning and releases).

Broader context for this repository: QuantEcon/meta#304 (toolkit proposal) and QuantEcon/meta#335 (benchmarking programme).