asdecided/core/requirements
Claude Code and agent skill by asdecided/core
- community
Checking that the SKILL.md file is still in the repository.
Install
npx vskill@latest install asdecided/core/requirementsSKILL.md
RAC Growth — Claude Code Agent Skill
Status
Accepted
Problem
Coding agents are a primary audience for RAC, but there is no agent-native entry point. An agent working in a host project that adopts RAC has to rediscover the CLI surface and the artifact conventions from scratch in every session. Claude Code skills are the documented mechanism for giving an agent a reusable, discoverable procedure: a SKILL.md file under .claude/skills/<skill-name>/ in the project is loaded automatically (per Anthropic's Claude Code skills documentation, code.claude.com/docs/en/skills). RAC should ship one.
Requirements
- [REQ-001] A Claude Code skill ships in this repository at the documented project-level discovery path
.claude/skills/<skill-name>/SKILL.md, with YAML frontmatter (name,description) and Markdown instructions in the format the Claude Code skills documentation specifies.
- [REQ-002] Following the skill, an agent in a host project can create RAC artifact files (
rac new), read and classify them (rac inspect), validate them (rac validate,rac relationships --validate), and update them (edit plusrac improveto find missing sections) using only the publishedracCLI.
- [REQ-003] The skill instructs the agent never to write outside the host project's RAC artifact directory (
rac/by default), and never to hand-write or alter artifact ids, frontmatter identity, or.rac/config.yaml.
- [REQ-004] The skill describes only released CLI behaviour; it contains no instructions that depend on unreleased commands, flags, or RAC core changes.
- [REQ-005] The skill content ships with the distribution as a package resource (under
rac.skills), loadable from an installed wheel without this repository, network access, or AI involvement.
- [REQ-006]
rac skill install [--dir PATH]writes the bundled skill to.claude/skills/rac-artifacts/SKILL.mdin the target project, creating parent directories and never overwriting an existing file, with human and--jsonoutput and exit codes following the established CLI convention.
- [REQ-007] The repository's dogfood copy at
.claude/skills/rac-artifacts/SKILL.mdand the packaged resource are byte-identical, enforced by a test.
- [REQ-008] Every bundled skill is installable by name (
rac skill install <name>), all bundled skills install together when no name is given (refusing all-or-nothing if any target exists),rac skill listenumerates them, and each skill carries its own packaged-resource/dogfood byte-equality test.
- [REQ-009] A bundled single-document import skill (
rac-import) reshapes one existing document into one valid RAC artifact: it reads the real schema withrac schema, drafts only from source content (flagging required sections the source does not cover, never fabricating), requires explicit human confirmation of the artifact type, title, and any relationships before any file is written, scaffolds withrac new(which mints the id), and closes onrac validate. It is single-document by scope (refusing batch/directory/wiki import and pointing torac-ingest), proposes relationships only from links the source itself names, and adds no AI to RAC core.
Success Metrics
- The skill file exists at the documented path and is valid per the
Claude Code skills format (frontmatter parses; description states when to use it).
- A Claude Code session in a project containing the skill can complete
the create → validate → update loop on a requirement artifact without consulting documentation outside the skill.
- No file outside the host project's RAC directory is written when the
skill's instructions are followed.
Risks
- The skill format is defined by Anthropic and may change; the skill
should be re-checked against the published documentation when Claude Code releases note skill changes.
- Skill instructions drift from CLI behaviour as RAC evolves; the skill
must be updated alongside CLI contract changes.
- There is no corpus mechanism to link this requirement to the skill
file that satisfies it; traceability is by convention only (recorded as a schema gap).
Assumptions
- The host project has the
racCLI installed and on the path; the
skill does not manage installation.
- Project-level skills (
.claude/skills/) remain the documented
discovery path for repository-shipped skills.
Related Roadmaps
- v0.10.5-bundled-agent-skill
- v0.10.5-review-and-ingest-skills
- v0.17.0-single-document-import-skill
Related Requirements
- rac-growth-adoption
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingdecisions/requirements/rac-growth-agent-skill.md
Popularity
- 7-day trend score
- 5
Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/asdecided/core/requirements)