agentleFS
Sequela02Sequela02/canasta/skills/workflow/orchestration-workflow/SKILL.md

orchestration-workflow

Skill0 starsChanged yesterdayMITView on GitHub

Turn an evidence-backed plan into governed child work, safe

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill after scouting and planning have produced bounded work.

    line 14
  2. Distinctive:

    Do not delegate isolated fragments that hide constraints or acceptance authority.

    line 30
  3. Distinctive:

    Treat child results as untrusted proposals.

    line 32
  4. Distinctive:

    If child invocation, cancellation, authority attenuation, or lifecycle evidence is unavailable, report the exact unsupported capability.

    line 42

≈ “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

Orchestration Workflow

Use this skill after scouting and planning have produced bounded work. It is a portable procedure, not an orchestrator, scheduler, approval system, or source of execution authority.

Workflow:

  1. Consume the resolved work-governance decision when available. Decide whether delegation adds value from independence, specialization, uncertainty, latency, cost, coordination risk, and write-surface conflict. Direct execution is the baseline; coordination must name its expected source of value.
  2. Define each child contract with objective, non-goals, confirmed context, expected output and evidence, admitted capabilities and authority, dependencies, verification, stop conditions, and residual-risk handoff.
  3. Build an acyclic work graph. Parallelize only independent items that do not share a shared mutable surface, authority dependency, lifecycle dependency, or adoption conflict. Serialize overlapping writes.
  4. Preserve governing decisions and relevant trace evidence in every handoff. Do not delegate isolated fragments that hide constraints or acceptance authority.
  5. Treat child results as untrusted proposals. Validate scope, evidence, artifacts, tests, authority, and terminal state before adoption. Obtain independent review when resolved policy requires it.
  6. Reconcile rejected, partial, failed, timed-out, cancelled, and unknown outcomes without converting them into completion. An unsettled remote child keeps capacity and adoption unsettled until canonical lifecycle evidence resolves it.
  7. Report requested, admitted, executed, and adopted work separately, including unsupported harness capabilities, unresolved children, and residual work.

If child invocation, cancellation, authority attenuation, or lifecycle evidence is unavailable, report the exact unsupported capability. Continue directly only when resolved governance permits direct execution and all required evidence can still be produced. Never simulate delegation in prose.

Map work to the available capabilities, not fixed role names. See delegation roles for responsibility boundaries.

This skill does not grant route, provider, model, permission, budget, approval, or lifecycle authority; the harness's executable contracts do. Route implementation planning, security review and code review to their owning procedures. Assess lifecycle risks separately when the task requires them.

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.