Docs/Skills & MCP tools

flowrail-lineage

live

Renders the full event chain for a design review as an ASCII tree, inside Claude Code. The agent calls it when it needs to know what's already decided. You call it for a quick sanity check without leaving the terminal.

01

When it runs

On demand. Claude invokes it when it needs context, typically after a verify failure or before suggesting a refactor that might overlap with an earlier decision. You can also ask Claude to render the tree directly.

What you see in Claude Codev0.4.2
↳ flowrail-lineage  dr_7f3a9e2c

dr_7f3a9e2c  password-reset.spec.md
├── 0.1s    review opened              4 threats predicted
├── 38.1s   verify denied              src/auth/reset.ts  · no-secrets-in-code
├── 38.6s   verify allowed             src/auth/reset.ts
├── 1m04s   dep-check allowed          npm install argon2
└── 4m12s   review resolved            clean · merged
MCP toolv0.4.2
flowrail_get_lineage
Request / responsev0.4.2
// request
{ "design_review_id": "dr_7f3a9e2c" }

// response
{
  "tree":        "<ASCII tree as one string>",
  "event_count": 5,
  "duration_ms": 252000
}

The tree is rendered server-side from the events table; the skill returns one string. For richer drill-downs (per-guardrail summaries, per-package signals) use the dashboard at /reviews/<dr_id>. The skill is for in-context retrieval. The dashboard is for reading.