Suede AI Code Skills Every merge gets a verdict it can't argue with.
One run returns both halves of a review: severity-ranked findings with file and line evidence, and the A to F ship grade where the weakest of seven lanes caps the letter. A hardcoded secret, or an auth check a request param can walk around, locks the grade at F and prints the file and the line. It runs only when you ask for a review, a grade, a security check, or a merge verdict.
"grade this diff before I merge"->/suede-codeAsk in plain words. The router reads the request and loads this lane; MCP agents find it with list_suede_skills.
claude code · sample session
$ /suede-code review src/auth/login.ts7 lanes armed · instant-F triggers liveP1 · session token written to logs · login.ts:88SHIP GRADE: D · HOLDfix brief attached · re-run to clear
Deep, evidence-based code findings and a blunt A-F ship grade in one pass.
File-level evidence, impact, and a concrete fix path for every blocking finding.
Target, intent, and risk-lane mapping before judging a diff, PR, route, API, or release.
Repo gates such as typecheck, lint, tests, dependency audit, and secret scanning when they exist.
Instant-F triggers, language traps, OWASP checks, grade caps, and release-review modes.
A deploy safety gate that turns the grade into ship, ship-with-caveats, or hold.
When to use it
Use it for explicit review, grade, audit, security-check, or ship-gate requests.
Use it on diffs, PRs, files, route changes, API changes, release builds, and fix verification.
Use deeper review for auth, payments, migrations, public APIs, and cross-file behavior.
Not for: wiring CI or branch protection (use suede-ci-gate).
Best prompt
Use $suede-code to run a --standard review of the current diff against main. Intent: add checkout webhook handling. Risk lanes: backend, payments, data, auth, tests, and deploy readiness. Run available gates, report findings with file evidence, grade each lane, apply the deploy safety gate, and give the final ship gate.
Related skills
Related pages: Suede Ship Gate makes CI enforce the verdict, and Suede AI Eval covers LLM and agent behavior the diff ships.
Install this skill
It ships inside the Suede AI Skills pack, so one install brings the whole pack and this skill with it. Claude Code takes two commands the first time and one after that; Codex installs the same pack natively.