Skill v1.0.0

metadata: version: "1.0.0"

name: "gate-enforcement" description: "Run risk-based ai-agent-board quality gates after code-producing work." domain: "testing" confidence: "medium" source: "ai-agent-board gate enforcement and commands"

Context

Use this skill after any task that writes code or changes runtime behavior. Documentation-only Squad state changes only need consistency validation unless they add scripts or app-facing behavior.

Patterns

• Identify the affected surface in .squad/coverage-matrix.md.
• For code-producing work, run the deterministic required gate: npm run gate:required.
• The required gate runs npm run build:client, npm run build:server, and npm run test:e2e:required; do not replace it with ad hoc build or E2E commands.
• Install the local pre-push hook with npm run hooks:install so Git uses .githooks/pre-push.
• Prefer focused existing tests before broad suites, but required gate evidence must still include npm run gate:required.
• E2E portability/setup problems, including hard-coded Unix paths such as /tmp/test-repo, are blockers to fix in the test setup before approval; they are not acceptable reasons to defer affected E2E coverage.
• Fix-loop ceiling: 3 attempts or 5 minutes on the same failing class, then escalate to Morpheus/Switch.

Examples

• Agent runtime change: Neo owns implementation; Switch requests focused runtime tests plus npm run gate:required.
• Board UI change: Trinity owns implementation; Switch requests focused e2e/component evidence plus npm run gate:required.

Anti-Patterns

• Do not weaken or skip tests to make a gate pass.
• Do not treat unrelated baseline failures as proof the current change is safe.
• Do not skip affected e2e for convenience, portability, unavailable dev servers, or setup friction; start/fix the required processes or fail the gate.
• Do not replace required e2e evidence with a build-only claim for browser workflow or client-server behavior changes.
• Do not run softer ad hoc E2E commands when npm run gate:required is available.