agentleFS
Sign inStart free
aishwaryajayanth1820-gitaishwaryajayanth1820-git/code_graph_agentic_observable_coding/.claude/skills/orchestrate-cycle/SKILL.md

orchestrate-cycle

Skill1 starsChanged 32 days agoApache-2.0View on GitHub

Orchestrator skill. Runs the code → review → refine cycle over a locked code-graph design - slices ready nodes into parallel-safe tasks, dispatches coder agents with task briefs, resolves contract-change escalations through ADRs, runs the reviewer agents, gates EVALUATED, and measures EVAL adherence and platform maturity to improve prompts. Use once the C4 design is locked and until the project is complete.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Review: When a batch's nodes are TESTED, run the reviewers in parallel, as described in reference/review-cycle.md.

    line 14
  2. Very specific:

    Measure and evolve: After each cycle, run reference/maturity-review.md: cg metrics, cg eval report, cg check.

    line 16
  3. Distinctive:

    Keep coders on their nodes.

    line 23

≈ “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

Orchestrate the build cycle

Tool: python <project>/code_graph/cg.py ... --actor orchestrator. Agents: coder, arch-reviewer, app-design-reviewer, code-reviewer, security-reviewer, performance-reviewer.

One cycle

  1. Slice. From cg node list --status PLANNED, pick nodes whose dependencies are done: platform and interfaces before the use cases that plug into them. Group them into tasks with no shared nodes, so they are safe to run in parallel.
  2. Create and dispatch. For each task, run cg task new --title --desc --nodes --agent coder-N, then launch a coder agent with a brief built from reference/task-brief-template.md. Launch parallel coders in one message.
  3. Escalations. A coder that returns BLOCKED with an ADR id goes through reference/escalation-protocol.md. Accepting an ADR means notifying every agent in impacted before you resume anyone.
  4. Review. When a batch's nodes are TESTED, run the reviewers in parallel, as described in reference/review-cycle.md.
  5. Refine or gate. FAIL verdicts become REFINEMENT tasks for coders; re-review only the failed EVALs. When every latest verdict is PASS or WAIVED, run cg node status <id> EVALUATED.
  6. Measure and evolve. After each cycle, run reference/maturity-review.md: cg metrics, cg eval report, cg check. Append to <project>/design/evolution_log.md, and move the guidance behind the least-followed EVALs into .claude/playbooks/coder-lessons.md.

Repeat until every node is EVALUATED. Report progress to the user with the UI link (#view=metrics).

Rules for yourself

  • The graph is the only status board. Never track progress anywhere it cannot see.
  • Accept an ADR yourself only when it stays inside the approved design. Ask the user when it changes a platform contract or the component structure.
  • Keep coders on their nodes. A cross-cutting fix is a new task, not scope creep.

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.

Check your setup, free