@flowrail/skills
v0.0.2
Published
FlowRail skill markdown for Claude Code agents (POC). Auto-invocable skills for design review, verify, lineage, and dep-check; copied into .claude/skills/ by @flowrail/init-poc.
Readme
@flowrail/skills
Source of truth for FlowRail's four Claude Code skills. Each skill is a thin agent-facing pointer that calls one MCP tool on the FlowRail server:
| Skill | When the agent uses it | MCP tool |
|---|---|---|
| flowrail-design-review | Tester wrote a *.spec.md or asked for a threat model | flowrail_design_review |
| flowrail-verify | Hook auto-fired on a code Write OR tester asked "why was this blocked?" | flowrail_verify_code |
| flowrail-dep-check | Hook auto-fired on npm install OR tester asked "is X@Y safe?" | flowrail_check_dep_install |
| flowrail-lineage | Tester asked "what did FlowRail decide for this task?" | flowrail_get_lineage |
Ships:
manifest.json— single source of truth for the four skill names. Imported by@flowrail/init-poc(the installer) and the skills package's own integrity tests.flowrail-*/SKILL.md— the agent-facing instructions. Copied into.claude/skills/by@flowrail/init-poc.
You don't install this directly — @flowrail/init-poc pulls it as a dep and copies the markdown into your repo.
Source + docs: https://github.com/anshumanbh/flowstate
POC, eval-only.
