Agent-Skills.md

Agent Skills: Verifying contracts-engine changes

Drive the contracts engine's runtime surface (the stdio MCP server) to verify changes end-to-end. Scoped to public-plugins/plugins/healthcare/skills/contracts/ and ../../servers/documents/.

UncategorizedID: anthropics/healthcare/verify

Repository

anthropics/healthcare
anthropics
33981

Install this agent skill to your local

pnpm dlx add-skill https://github.com/anthropics/healthcare/tree/HEAD/plugins/healthcare/skills/contracts/.claude/skills/verify

Skill Files

Browse the full folder contents for verify.

Download Skill

Loading file tree…

plugins/healthcare/skills/contracts/.claude/skills/verify/SKILL.md

Skill Metadata

Name
verify
Description
Drive the contracts engine's runtime surface (the stdio MCP server) to verify changes end-to-end. Scoped to public-plugins/plugins/healthcare/skills/contracts/ and ../../servers/documents/.

Verifying contracts-engine changes

The engine is the bundled MCP server at ../../servers/documents.mjs (source ../../servers/documents/src/). Most changes are drivable without a full /contracts session by speaking JSON-RPC to the server over stdio.

  • Build: cd ../../servers/documents && bun install && bun run build && bun run bundle (typecheck + regenerate the committed bundle — always rebundle before testing; the plugin runs the bundle, not src/).
  • Data: ~/.claude/data/healthcare/contracts/data.sqlite (shared across checkouts; safe to delete — schema v-check tells users to do the same).
  • Tool schemas must be JSON Schema draft 2020-12. The API validates them when an agent spawns, so a bad schema shows up as "agent terminated early: input_schema is invalid" — never as a server error, and never in a plain tools/list. Two zod idioms silently emit draft-07: z.tuple([...]) (array-form items → use z.array().length(n)) and z.union([...])/.nullish() on primitives (type: [...] → use .optional() or one type). servers/documents/test/schema.test.ts guards this; run bun test after touching any tool's inputSchema, and spawn a real agent (claude --plugin-dir <plugin> -p "spawn subagent_type 'healthcare:contracts-reader' …" --allowedTools Agent) after changing agent tool lists.
  • MCP smoke: pipe initialize → notifications/initialized → tools/list / tools/call lines into node ../../servers/documents.mjs and read the JSON-RPC replies. A sequential client (write line, read reply) avoids out-of-order confusion. Core chain worth driving after engine changes: corpus_register (temp dir with a .txt) → ingest → write runs/briefs → find with a verbatim quote (expect kind:"exact") AND a bogus quote (expect the cite error) → export_report.
  • Sweep fan-out is parallel Agent calls in one message (10-wide, getMaxToolUseConcurrency), deliberately NOT a Workflow — a workflow caps agents at min(16, cores−2), which is 2 on a cloud container and serializes the sweep. To check parallelism after a real run: sql "SELECT worker, min(created_at), max(created_at) FROM findings WHERE run_id='<id>' GROUP BY worker" — the windows should overlap, not chain end-to-start.
  • SKILL.md / agents/contracts-conductor.md / steps/*.md prose changes: no cheap harness — check internal consistency (tool names match the server's tools/list, step file paths exist) and, for protocol changes, drive at least the mechanical tool calls the prose mandates exactly as written.