Author and validate Plasm domain models (CGS, `domain.yaml`), capability mappings (CML, `mappings.yaml`: HTTP REST, GraphQL via `transport: graphql`, composed reads via `transport: view`), and `views:` DAGs. Test interactively with `plasm-repl` against Hermit mocks or live/sandbox backends. Use when extracting schemas from OpenAPI specs, writing or editing `domain.yaml` / `mappings.yaml`, validating mappings, iteratively developing typed agent tooling, or driving Cursor / Claude / Codex / other coding agents — point them at this skill as the full playbook; do not duplicate per-API runbooks into agent prompts.
Polish an existing Plasm catalog to be publish-ready. Runs an autonomous diagnostic / fix loop on `apis/<api>/`: CGS validation, mapping checks against the OpenAPI spec when available, teaching table prompt hygiene, descriptions and discovery audit, eval coverage, and transport evidence via the e2e-test skill. Iterates until the catalog passes its quality bar, then reports a before / after delta. Use after `plasm-authoring`, `plasm-catalog-reprint`, or any time a catalog is "almost there but not shipped". Trigger phrases: "polish the catalog", "clean up apis/<api>", "make it ship-ready", "fix the descriptions", "fix the eval coverage", "polish".
Score a Plasm catalog against the canonical rubric and produce a scorecard. Compares semantic compression, typed values, relation utility, action outputs, mappings completeness, view usage, transport evidence, eval coverage, and README quality. Use after `plasm-catalog-polish` (or after a fresh authoring pass) to grade catalog quality, compare two catalogs, or set a target before promoting a catalog to "trusted" status. Trigger phrases: "score the catalog", "scorecard", "grade apis/<api>", "compare apis/<a> vs apis/<b>", "is this catalog ship-ready", "rate this catalog".
Operational end-to-end testing for a Plasm catalog. Use after authoring or editing `apis/<api>/`, before publishing or scoring, or whenever a catalog has not been transport-tested in its current shape. Runs Hermit-backed mock tests when the README or source docs identify an OpenAPI spec, then live or vendor-sandbox tests when credentials and sandbox endpoints are available. Records what ran, what was skipped, and why. Trigger phrases: "e2e test", "transport test", "Hermit test", "smoke test the catalog", "verify the mappings against the spec", "drive plasm-repl against the mock", "live test the catalog", "sandbox test".
Full-cutover regeneration of a weak Plasm catalog. Use when a catalog's CGS structure is fundamentally wrong (RPC-shaped capabilities, missing entities, bad relation choices, or accumulated debt that polish cannot fix). Preserves useful prior evidence (eval cases worth keeping, vendor doc references, README scope) but replaces `domain.yaml` and `mappings.yaml` with a fresh authoring pass under the current `plasm-authoring` skill. Trigger phrases: "reprint the catalog", "regenerate apis/<api>", "this catalog is wrong", "start over on apis/<api>", "redo the model", "full rewrite".
Post-authoring retrospective on a Plasm catalog run. Captures systemic improvements to the Plasm authoring pipeline itself — skill text, reference doc, fixtures, validators, runtime expressiveness, eval harness — so the next catalog comes out better. Emits a structured findings record and, when actionable, a GitHub issue draft against `plasm-core`. Use after `plasm-authoring`, `plasm-catalog-polish`, `plasm-catalog-reprint`, or `plasm-catalog-score` when the session surfaced something that should change the press, not just the catalog. Trigger phrases: "retro", "retrospective", "post-mortem on this catalog", "what should we change about Plasm", "file a retro", "improve the press", "lessons learned".
Redirect — canonical CGS / CML authoring skill lives under `skills/plasm-authoring/`.
Redirect — canonical skill lives under `skills/plasm-catalog-e2e-test/`.