namba-ai: Skill for Claude Code

State effect: mutating workflow entry point. Use help/probe paths read-only, and otherwise expect repository state or GitHub state to change.

Generated instruction contract for this command skill:

• Purpose: keep the role or command scope explicit, bounded, and testable.
• Boundary: honor read-only versus mutating state effects, configured sandbox mode, and assigned file or workflow ownership.
• Required output: report concrete actions, changed paths or artifacts, validation evidence, and pass/fail status or blockers.
• Pass/fail criteria: claim success only when acceptance criteria and configured validation are satisfied; otherwise name the exact blocker and impact.
• Evidence expectations: cite source artifacts such as SPEC files, .namba/ configs, diffs, test output, PR/check links, or generated manifests instead of relying on unsupported assertions.
• Security responsibilities: never expose or commit secrets; treat auth, privacy, destructive commands, permission changes, and external network or credential use as security-sensitive.
• Destructive command and escalation policy: do not run destructive commands unless explicitly requested; request approval for privileged, networked, or sandbox-blocked actions only when the active approval mode allows it, and otherwise report the blocker or use a safe non-escalating path.
• Fallback implementer boundary: if a specialist path is unavailable and the main/default implementer takes over, stay within the assigned scope and preserve the same evidence and validation duties.
• Portability: keep durable guidance non-project-specific unless the current repository config or SPEC explicitly provides the project detail.

Use this skill when the user explicitly says $namba-run, namba run SPEC-XXX, or asks to execute a SPEC through Namba.

Behavior:

• Read .namba/specs/<SPEC>/spec.md, plan.md, and acceptance.md before implementation.
• Read .namba/specs/<SPEC>/reviews/readiness.md when it exists so advisory review depth is visible before coding starts.
• Read .namba/specs/<SPEC>/frontend-brief.md when it exists; it is the canonical source for frontend task classification and gate state.
• In an interactive Codex session, prefer Codex-native in-session execution over recursively calling namba run.
• Only use the standalone CLI runner for --solo, --team, --parallel, --dry-run, or when the user explicitly wants the non-interactive runner path.
• For --solo, stay inside one runner unless one domain clearly dominates and a single specialist would materially reduce risk.
• For --team, prefer one specialist when one domain dominates, expand to two or three only when acceptance spans multiple domains, and keep one integrator plus final validation owner in the workspace.
• For --team, honor each selected role's model and model_reasoning_effort metadata from .codex/agents/*.toml so planner/reviewer/security roles can think harder without making every delivery role heavy.
• Route art direction, palette/tone logic, composition, motion intent, Figma critique, and generic-section redesign work to namba-designer; route component boundaries, state ownership, and UI delivery planning to namba-frontend-architect; route approved UI implementation to namba-frontend-implementer; route mobile-specific UI delivery to namba-mobile-engineer; route API, schema, and pipeline work to backend/data; route auth, secrets, and compliance work to security; route deployment and runtime work to devops.
• frontend-major work must not move into architecture or implementation until frontend-brief.md shows coherent problem, reference, critique, decision, prototype evidence, a complete Do-Not Design Contract with reference-driven asset manifest, generated-image decision fields, generation plan, and aligned design clearance; frontend-minor keeps the lightweight advisory path.
• Treat review readiness as advisory by default for non-frontend and frontend-minor work, but block explicit frontend-major execution when the frontend brief is missing required evidence, internally contradictory, mismatched with design-review summaries, or missing/insufficient negative-first contract evidence.
• For frontend-major, first frontend and first major screen phases default to Asset mode: generated-images and Imagegen requirement: required; accept existing-assets or not-applicable only with validator-readable Asset decision proof.
• For frontend-major implementation results, require a Do-Not Design Violation Check that cites changed files, names any banned pattern found, and cites the exception path when one is used; when Imagegen requirement: required is present, require Generated Asset Evidence with manifest path and repeatable Asset ID blocks containing generated file, saved asset path, prompt summary, intended UI usage, and rendered usage evidence.
• For browser-rendered frontend work, use managed server lifecycle, wait for rendered DOM state, capture screenshots, inspect console errors, and prefer Playwright checks when the surface runs in a browser.
• Run validation commands from .namba/config/sections/quality.yaml and finish with namba sync. Use namba pr and namba land for the GitHub handoff and merge cycle instead of overloading sync.
• Collaboration defaults: branch from main, open the PR into main, write the PR in Korean, and request Codex review only when namba pr --review or queue --review is explicit; use @codex review as the request command.