Curated conformance corpus for agent-cards-spec v0.1 consumers. Final piece of the test-vectors corpus pattern across all 4 Kinetic Gain Suite specs.
Status: v0.1.0 — 10 vectors across 3 categories. Node 20/22 supported.
| Spec | Test vectors |
|---|---|
otel-genai-test-vectors |
OTel GenAI semantic conventions |
prompt-provenance-test-vectors |
prompt-provenance-spec v0.1 |
evidence-bundle-test-vectors |
evidence-bundle-spec v0.1 |
agent-card-test-vectors ← this repo |
agent-cards-spec v0.1 |
| Category | Count | Description |
|---|---|---|
valid |
4 | One vector per autonomy level (assistive, supervised, autonomous) plus a minimal-required floor. |
dirty |
4 | One spec rule violation per vector — pinned in expectedFindings. |
shape |
2 | with-all-optional (every optional populated) + no-tools-no-evals (pure-text reasoning agent). |
import { listManifests, loadVector, loadVectors } from "agent-card-test-vectors";
// Pick the dirty one for a specific rule
const v = loadVector("autonomous-without-iru");
console.log(v.manifest.expectedFindings); // ["autonomous-without-incident-response-uri"]
// Sweep
for (const v of loadVectors({ category: "dirty" })) {
const findings = myValidator(v.payload).map((f) => f.code);
for (const expected of v.manifest.expectedFindings ?? []) {
assert(findings.includes(expected), `${v.manifest.id} → expected ${expected}`);
}
}agent-cards-spec— the schema these vectors conform to (or deliberately violate).agent-card-diff— diff two cards across versions.agent-card-fleet-summary— analyze a directory of cards.agent-card-tool-coverage— verify declared tools against an MCP server.kg-validate-action— per-doc CI validator.
npm install
npm run lint && npm run typecheck && npm run coverage && npm run build
npm run demo