Cursor agent skills

Cursor Agent Skills for documentation hygiene, planning, security passes, private-to-public publishing, and agent alignment. Each pack lives under .cursor/skills/<name>/ with a SKILL.md (and optional REFERENCE.md). MIT licensed; some packs trace to Matt Pocock's skills (see Acknowledgements).

Quick start

Recommended — /setup-eunai-skills: Copy the setup-eunai-skills folder into your repo at .cursor/skills/setup-eunai-skills/, then ask your agent to run /setup-eunai-skills (or describe that intent in chat). The skill installs the canonical write-a-skill helper, recommends packs for your repo, and authors adapted copies.

The orchestrator lists available packs from the public mirror eunai/skills (.cursor/skills/ on main). If that listing is empty or you are offline, use the manual path below.

Fallback — manual copy: Copy any folder from .cursor/skills/ into your project at .cursor/skills/<same-folder-name>/. Cursor loads project skills from there. Open each pack's SKILL.md for triggers and behavior.

Reference

One-line capability blurbs are the first sentence of each pack's YAML description: (see SKILL.md frontmatter). Groups are ordered bootstrap → docs → design → authoring → output style; packs are alphabetical within each group.

Bootstrap & repo scaffolding

• bootstrap-agent-index — Creates or updates a maintainer-facing agent_index.md that is the canonical start-here entrypoint for AI agents and contributors.
• bootstrap-agent-playbooks — Creates or updates an agents/ folder of short topic playbooks for maintainers and AI agents.
• bootstrap-modules-readme — Creates or updates modules/README.md describing package boundaries, allowed dependency edges, and forbidden edges.
• bootstrap-private-public-mirror — Sets up a private repo to publish a sanitized public mirror via git filter-repo, including docs split, public stubs, publisher script, whitelist doc, link repair, and dry-run validation.
• setup-eunai-skills — Bootstraps the eunai skill set into a target Cursor repo by installing the canonical write-a-skill helper (eunai-first, Pocock fallback with Cursor adaptation), recommending skills based on the target repo's language and layout, then authoring each confirmed skill via /write-a-skill with adaptations.

Doc & process discipline

• adr-record — Creates append-only Architectural Decision Records with stable numbering and supersedes chains.
• changelog-discipline — Maintains CHANGELOG.md and optional public changelog stubs using Keep a Changelog and Semantic Versioning.
• docs-drift-audit — Compares repo facts against documentation and emits a structured drift report, read-only by default.
• maintain-recommended-libraries — Creates or updates a decision-oriented recommended_libraries.md memo of library candidates and trade-offs.
• project-status-slice-report — Creates and updates a slice-based project_status.md plus per-slice slice_report/ entries with required verification evidence. Roadmap diagrams use a fixed flowchart LR Mermaid house style with edge lines capped at four --> per source line and bridge-node continuations (see the pack’s REFERENCE.md).
• security-review-pass — Produces dated internal security review reports with severity-tagged findings and tracker checklists.
• vertical-slice-planner — Turns an issue, PRD, or rough idea into thin independently shippable vertical slices in a Markdown plan.

Design conversation

• grill-me — Guides the agent to stress-test a plan or design through relentless, one-question-at-a-time interviews until decisions align.
• grill-with-docs — Runs a documentation-aware grilling session that challenges a plan against domain language and recorded decisions, updating CONTEXT and ADRs as terms crystallize.

Skill authoring

• revise-prompt — Restructures a draft into a structured, agent-readable prompt with a stable core scaffold and optional sections, then delivers it in one copy-paste code fence plus a short change summary, optionally ending with a one-line Y/N offer to execute the revised prompt next.
• write-a-skill — Guides creation of Cursor Agent Skills under .cursor/skills/ with proper structure, progressive disclosure, and bundled resources.

Agent output style

• caveman — Configures the agent for ultra-compressed replies that drop filler while preserving full technical accuracy.

Acknowledgements

This collection is inspired by Matt Pocock's skills (MIT). Packs derived from that work — caveman, grill-me, grill-with-docs, and write-a-skill — carry a Pocock source line and a note that they were adapted for Cursor in each SKILL.md footer. Other packs were authored here and carry an eunai/skills (MIT) footer. Per-skill footers are the detailed source of truth.

License

This project is licensed under the MIT License.

Changelog

See CHANGELOG.md for release notes.