agentleFS
vosslabvosslab/vosslab-skills/skills/management/gas-town-workflow/SKILL.md

gas-town-workflow

Skill5 starsChanged 23 days agoMITView on GitHub

Coordinate agents with Gas Town role mapping and convoy-based task decomposition. Use only when the user explicitly requests Gas Town, convoy workflow, theatrical roles, or this repo's agent-role system; not for generic parallel work.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Use this only for an explicit Gas Town request.

    line 8
  2. Distinctive:

    Read references/glossary.md when Gas Town terms or their tool equivalents matter.

    line 23
  3. Distinctive:

    Use the repository role catalog for fuller authority and escalation detail.

    line 43
  4. Very specific:

    Create one parent task for the initiative and atomic child tasks with role tags such as [CODER] or [TESTER].

    line 49
  5. Very specific:

    Put Role, Convoy, and measurable Done when criteria in every task.

    line 51
  6. Distinctive:

    Use direct messages for a specific escalation, coordination need, or completion signal.

    line 61

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Gas Town workflow

Use this only for an explicit Gas Town request. For ordinary approved-plan execution, use delegate-manager-to-subagents; do not silently replace its plain manager/coder/reviewer vocabulary with theatrical terms.

Operating contract

  • GUPP: an assigned, unblocked task starts immediately.
  • MEOW: decompose a goal into atomic, independently completable tasks with objective done-when criteria.
  • NDI: preserve task state, detect failed or stalled work, retry or escalate it, and never silently drop it.
  • A convoy is the grouped, dependency-ordered set of those tasks. Every task has one owner, an explicit role, a convoy name, a status, and a completion signal.

Read references/glossary.md when Gas Town terms or their tool equivalents matter. Read references/convoy-templates.md before creating a convoy.

Route work by role

Gas Town roleAgentBoundary
CrewcoderProduction code and small documentation updates
RefineryintegratorMerge and conflict resolution
WitnessmonitorObserve progress and report stalls
DeaconschedulerTrigger retries; do not diagnose
DogsmaintainerCleanup, lint, and index regeneration
--reviewerRead-only code or plan review
--testerTests and validation
--architectCross-cutting design decisions
--plannerPlans and documentation only
--orchestratorParallel coordination

Use the repository role catalog for fuller authority and escalation detail. Route design conflicts to architect, plan ambiguity to planner, implementation failures revealed by tests to coder, and detected stalls to orchestrator.

Create and run a convoy

  1. Create one parent task for the initiative and atomic child tasks with role tags such as [CODER] or [TESTER].
  2. Put Role, Convoy, and measurable Done when criteria in every task.
  3. Set the dependency graph before assignment; give each task to its matching role and keep ownership exclusive.
  4. Each agent claims its highest-priority unblocked role-matching task, marks it in progress, performs the work, records evidence, and explicitly completes or escalates it.
  5. A monitor checks for genuine stalls; a scheduler retries work that has been diagnosed as retryable. At convoy close, verify every done-when criterion and shut down no-longer-needed teammates.

Use direct messages for a specific escalation, coordination need, or completion signal. Do not broadcast routine status. Dispatch a fresh subagent for each atomic task; status messages do not convert a finished task into a new editing assignment.

Escalation and completion

An agent outside its authority creates or updates the blocker, notifies the appropriate role, and continues with independent unblocked work. A completed task includes relevant test or review evidence, an updated task state, and a completion message to the lead. If no suitable task remains, report that fact to the lead rather than claiming another role's work.

For full templates, role vocabulary, and tool mapping, use the routed references above; they are intentionally outside this entrypoint's control plane.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.