Skip to content

KamiJeong/tech-bridge-book / .agents/skills/speckit-rules

speckit-rules

Verified

Codex Spec Kit multi-agent workflow rules and safety invariants.

Spec Kit Multi-Agent Rules

Core Order

Default to gated behavior. For a full epic, run:

  1. intake
  2. slicing
  3. scheduling
  4. per-slice specify
  5. clarify
  6. checklist
  7. plan
  8. tasks
  9. analyze
  10. implementation only after approval or explicit automation mode
  11. review
  12. integration
  13. token reporting
  14. commit only when explicitly allowed
  15. draft PR only when explicitly requested

Execution Modes

  • plan-only: run through analyze and token reporting; do not implement.
  • gated: run through analyze and ask before implementation, commit, or PR.
  • auto-implement: implement only low-risk slices after analyze passes; do not commit or open PRs unless explicitly permitted.
  • auto-commit: validate, token-report, and commit only when explicit permission is present.
  • auto-pr: implement, validate, token-report, commit, push, and create a draft PR only when explicitly requested.

auto-commit is an explicit permission flag, not a standalone mode. It may be enabled for gated or auto-implement after review and validation pass, and it authorizes the commit step inside an auto-pr run. Stacked PR publication is currently disabled; use a single draft PR.

Slicing And Scheduling

Treat a request as a slicing candidate when it spans multiple independent deliverables, shared contracts, or more than one implementation area, unless the user or workflow explicitly overrides the decision.

The scheduler must choose one of:

  • single-spec
  • sequential
  • foundation-first-then-parallel
  • parallel-safe
  • stop-and-ask

Choose stop-and-ask when dependency, conflict, or repository-state safety is unclear.

Stop Conditions

Stop on:

  • CRITICAL analyze issues
  • failed tests
  • high-risk slices without explicit approval
  • unrelated user changes before implementation, commit, or PR
  • unclear base branch
  • missing or unauthenticated GitHub CLI for PR mode
  • protected publishing branches: main, master, develop
  • explicitly configured blocking token budgets
  • missing token-analyzer in automation modes without approval

Issue Commentary

Always add a short issue comment before each phase and before any terminal action such as commit, push, or PR creation. The comment should note the scope, label change, and any gate result.

Token Observability

Token analysis runs after every completed phase. Missing token data must be reported. High token usage is not a quality gate unless explicit blocking thresholds are configured.

Path And Configuration Rules

Use project-local Codex paths:

  • .codex/agents/
  • .agents/skills/
  • AGENTS.md
  • .specify/token-analysis/

Preserve existing repository configuration unless a minimal update is necessary for this workflow. Do not assume Claude Code paths.

reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 469kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 321kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 321kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 321kUpdated Claude CodeCodexVerifiedInstall