justinedevs/meta-architect/build

Build

Use this skill inside Codex to convert the earlier review lanes into an implementation-ready decision.

Output

Produce:

• current readiness verdict
• blockers that still prevent implementation
• the narrowest viable build slice
• branch or worktree suggestions when relevant
• bounded execution state when the build lane is already in progress
• test and verification expectations
• repair path when the bounded slice needs another pass
• the exact next implementation step

Rules

• Do not claim readiness if architecture, evidence, logic, security, or DX/UX gaps remain unresolved.
• Keep the recommended build slice small, testable, and reversible.
• Treat $build as the sole owner of .ma/plans/build.md and build_status.
• If the user wants code immediately and the path is clear, end with a concrete implementation plan rather than more review prose.

Procedure

1. Read the current architecture, evidence, security, UX, and logic decisions.
2. Classify each unresolved item as a release blocker, a bounded implementation task, or accepted risk.
3. Select one narrow slice with explicit inputs, changed surfaces, acceptance checks, and rollback boundary.
4. Define the regression-first test shape and the exact verification commands.
5. Hand the bounded slice to execution and keep unrelated cleanup out of scope.

Default evidence dossier

Use `docs/reference/supply-chain-readiness.md` for package, publish, deploy, and artifact work. Identify provenance, verification, promotion, rollback, and tamper-detection evidence before calling a release ready. Do not claim an SLSA level without matching evidence.

Quality bar

• “Ready” means the slice is implementable and testable, not that every future feature is designed.
• No build step may silently publish, deploy, delete, or rewrite user-owned state.
• A blocked result includes the missing evidence and the smallest probe that can unblock it.

Example

For a plugin release, verify version alignment, package contents, endpoint health, imported skills, release notes, and rollback behavior before recommending publication.