Skip to content

rampstackco/claude-skills / skills/dispatch-authoring

dispatch-authoring

Verified

Write the structured handoff from a planning session to a coding agent: a dispatch with a header, Gate 0 checks, scope, specification, verification, delivery rules, and stop conditions, so the agent can execute unattended and return a reviewable draft PR. Use this skill whenever the user wants to hand work to an agent, write an agent brief, write instructions for Claude Code, Codex, or a Cursor agent, queue work to run overnight, or turn a plan into something an AI coder can execute without asking questions. Triggers on dispatch, hand off to an agent, agent brief, agent handoff, task brief for an AI coder, instructions for Claude Code, Codex task, Cursor agent, background agent, run this overnight, give this to an agent, unattended run. Also triggers when the user has a plan or a spec and asks how to get an agent to build it, even if they never say dispatch.

references/gate-zero-checklist.md

View on GitHub
# Gate 0 Checklist

Gate 0 is the set of checks an agent runs before it writes anything. Each check has a stop action. Use this list when writing a dispatch's Gate 0, and again when reviewing a dispatch before it goes out.

---

## Identity

- [ ] **Remote matches.** The agent compares the remote URL to the exact `owner/repo` in the header.
  Stop action: report the remote it found and write nothing.
- [ ] **Visibility matches.** The agent confirms public or private through the hosting platform, not by assumption.
  Stop action: report the mismatch. A private-intended change in a public repo is a leak.
- [ ] **Default branch named.** The dispatch names the default branch so the agent can refuse to touch it.
  Stop action: if the default branch is checked out with uncommitted changes, report and do not stash or reset.

## Isolation

- [ ] **Fresh worktree or branch.** The agent creates a new worktree (or a new branch in a disposable clone) for this dispatch only.
  Stop action: if the branch name already exists, report. Do not reuse or reset it.
- [ ] **Canonical checkout untouched.** The dispatch says where not to write.
  Stop action: any write outside the worktree is a stop.
- [ ] **One dispatch per worktree.** No other dispatch is running in the same worktree.

## Reading

- [ ] **Required reading listed in order.** Authoring guide, contribution guide, the closest existing example, then the scripts the dispatch calls.
- [ ] **Agent instruction files checked.** Any repo-level instructions for agents are read and followed.
- [ ] **Scripts read, not assumed.** The agent reads each script's docstring or header before running it.
  Stop action: if a named script does not exist, report. Do not write a replacement.

## Generated files

- [ ] **Generators identified.** For each generated file the work will change (manifests, lock files, catalogs, dist builds), the agent records the script and flag that produces it.
  Stop action: if no script produces it, report. Never hand-edit a generated file.
- [ ] **Generators runnable.** The agent confirms each generator runs on the clean tree before making changes, so later failures are attributable.
  Stop action: if a generator fails on the clean tree, report the failure before writing anything.

## Safety

- [ ] **No destructive operations needed.** The plan does not require force-push, history rewrite, deleting branches, or deleting data.
  Stop action: if it does, report and wait.
- [ ] **No secrets needed.** The work does not require credentials beyond what the session already has.
  Stop action: if it does, report. Do not search for credentials.
- [ ] **Public repo hygiene.** For a public repo, the dispatch lists what must not appear: internal paths, private repo names, client names, internal hostnames.

---

## Review questions for the planner

Before sending, read Gate 0 as the agent and ask:

1. Could the agent write to the wrong place if any one line here were wrong?
2. Does every check have an observable pass or fail?
3. Does every failure have a stop action?
4. Is any count or list hardcoded that the agent should derive at run time instead?
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: 471kUpdated 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: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated 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: 322kUpdated Claude CodeCodexVerifiedInstall