governance

name	governance
description	Use for governed software work that needs source-of-truth checks, routing, token/context governance, branch/PR discipline, validation honesty, and release readiness. Do not use as approval for writes, installs, migrations, broad plugin use, external tool activation, or global config changes.

Governance

Use this as the canonical final governance entrypoint for serious toolkit or project work.

This skill authorizes routing, planning, read-only checks, capability selection, context-budget discipline, safety gates, and validation gates only. It does not authorize writes, migrations, package changes, CI changes, MCP setup, global config changes, external installs, product-repository writes, or broad plugin/tool activation.

Operating Rules

• Verify repository source of truth before readiness, release, or implementation claims.
• Select the smallest useful profile, agent, skill, support-tool, and method surface.
• Classify GSD and Superpowers on every governance run as not needed, lens only, selected, invoked, or blocked-unavailable; include a one-line reason for each status.
• Treat GSD as required for serious multi-step work, audits, migrations, backend/database/security/SRE work, source adoption, release programs, and cross-repo/toolkit sync unless an explicit manual GSD-equivalent phase/state fallback is declared.
• Use GSD as not needed or lens only for tiny tasks where invoking it would create planning noise or unnecessary .planning/ churn.
• Use compact context packs for large tasks instead of whole-repo or whole-registry dumps.
• Keep selected agents separate from agents actually spawned.
• Treat registries, source records, dry-runs, reports, and metadata as evidence, not execution.
• Preserve WARN output in completion reports even when aggregate validation passes.
• Stop before security-sensitive, destructive, global, package, CI, or product-repository changes unless separately approved.
• Route normal-language tasks through .ai-toolkit/methods/governance/task-intake-routing-gate.md before coding. Canonical toolkit source: methods/governance/task-intake-routing-gate.md.
• Use .ai-toolkit/methods/governance/governance-lite-router-mode.md as a concise method-only governance mode for small/medium tasks; do not create or activate a governance-lite or router-lite skill. Canonical toolkit source: methods/governance/governance-lite-router-mode.md.
• Route serious implementation through .ai-toolkit/methods/reliability/coding-time-production-readiness.md when production-style coding-time gates are needed. Canonical toolkit source: methods/reliability/coding-time-production-readiness.md.
• Route unsafe-command questions through .ai-toolkit/methods/governance/agent-command-safety.md. Canonical toolkit source: methods/governance/agent-command-safety.md.
• Route package-manager or workspace migration questions through .ai-toolkit/methods/repo/package-manager-workspace-migration.md. Canonical toolkit source: methods/repo/package-manager-workspace-migration.md.
• Use .ai-toolkit/docs/PROJECT_TOOLING_OPERATING_MODEL.md for v0.2 project tooling boundaries. Canonical toolkit source: docs/PROJECT_TOOLING_OPERATING_MODEL.md.
• Use the v0.2.5 activation model: active-if-detected for existing project-owned tools, owner-approved-install for absent tools, ci-advisory before calibrated CI blocking, static-adopted for safe toolkit-owned static concepts, and forbidden-runtime for MCP/daemon/global/memory/watcher conflicts.
• Use templates/commit-message-template.md for governed release or hardening commits that need consistent why/what/validation evidence.
• Use .ai-toolkit/methods/internal/documentation-accuracy-guard.md when docs mention concrete symbols, commands, flags, routes, config keys, paths, examples, or behavior. Canonical toolkit source: methods/internal/documentation-accuracy-guard.md.

External Discipline Status Contract

Every governance report must include:

• GSD status: not needed, lens only, selected, invoked, or blocked-unavailable.
• Superpowers status: not needed, selected, invoked, or blocked-unavailable.
• Reason: a concise explanation tied to task size, risk, state tracking, planning, debugging, TDD, review, or verification need.

Do not claim GSD or Superpowers were invoked unless their workflows actually ran. Selected support tools, manual lenses, source records, registries, dry-runs, and planned follow-ups are not execution evidence.

Large-Task Context Pack

For large tasks, source reviews, implementation plans, PR reviews, or multi-agent handoffs, report:

• token mode: concise, standard, or detailed
• changed-file neighborhood selected and why
• compact agent context pack contents
• omitted context and reason
• source, method, and profile references used
• private-overlay, secret, credential, and product-repo exclusions
• context evidence label: project-map, manual/static, or tool-generated
• status values such as metadata-only or not invoked must be reported as status, not as context evidence labels

Completion Evidence

Report branch/baseline, selected lenses or agents, files changed, validation commands, skipped checks, WARN output, residual risk, and the next release action.

Also report GSD and Superpowers status for governed work, even when one or both were not needed.