agentleFS
larchankalarchanka/skills/agent-loop/SKILL.md

agent-loop

Skill31 starsChanged 7 days agoMITView on GitHub

Orchestrates a multi-agent loop involving planning, development, review, and publishing sub-agents to implement and publish features test-first and with high standards compliance.

  • Commits and pushes

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 8 edited versions we read keep these.

  • 6

    Distinctive

    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. Distinctive:

    Verify compile checks and execution results.

    line 46
  2. Very specific:

    Update the status of the task in tasks/todo.md to completed [x].

    line 47
  3. Very specific:

    Check tasks/todo.md for the next incomplete task.

    line 58
  4. Distinctive:

    If tasks remain, loop back to Step 2.

    line 59
  5. Distinctive:

    If all tasks are completed, proceed to Step 4.

    line 60
  6. Distinctive:

    Verify git status and stage all changed files.

    line 65
  7. Distinctive:

    Commit files using the Conventional Commits syntax and push to origin.

    line 66
  8. Very specific:

    All tasks in tasks/todo.md are marked [x].

    line 71
  9. Distinctive:

    All unit and integration tests in the test suite pass successfully.

    line 73

“Common” is measured: kept by at least 6 in 10 of the 8 edited versions of this file we could read. “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

Multi-Agent Iteration Loop

This skill coordinates four independent agents (planning, development, review, and publisher) in a continuous loop to deliver and publish spec-compliant, standard-abiding features.

graph TD
    Start["User Request / Spec"] --> Plan["1. Planning Subagent"]
    Plan --> SavePlan["Generate tasks/plan.md & todo.md"]
    SavePlan --> ApprPlan{"User Approval?"}
    ApprPlan -- No --> Plan
    ApprPlan -- Yes --> Dev["2. Development Subagent"]
    Dev --> ImpTask["Implement Next Task via TDD"]
    ImpTask --> Rev["3. Review Subagent"]
    Rev --> Check{"Any Findings?"}
    Check -- Yes (Bug / Smell) --> Dev
    Check -- No --> CheckMore{"More Tasks?"}
    CheckMore -- Yes --> Dev
    CheckMore -- No --> Pub["4. Publisher Subagent"]
    Pub --> OpenPR["Create Branch, Commit & gh pr create"]
    OpenPR --> Done["Done - Pull Request Published"]

Workflow Execution Steps

[!IMPORTANT] Coordinator Constraint: The coordinator agent executing this skill MUST NOT write implementation code, create source files, or modify files in the codebase directly during the development phase. The coordinator is solely responsible for planning, orchestrating the loop, conducting code reviews, and managing publishing. All code changes and task implementations MUST be delegated to the spawned subagents.

1. Planning Phase

  • Spawn the planning subagent to read the user request/specification and codebase structure.
  • The planning agent will output tasks/plan.md (architecture, design, tasks) and tasks/todo.md (checkbox todo list).
  • Once the plan is saved, present it to the user for feedback and approval.

2. Development Phase (Task Implementation)

  • Select the first incomplete task from tasks/todo.md.
  • Spawn the development subagent to implement that task.
  • The development agent must:
    • Operate incrementally, writing a failing unit test first.
    • Implement only enough code to pass the test.
    • Verify compile checks and execution results.
    • Update the status of the task in tasks/todo.md to completed [x].

3. Review Phase (Quality Gates)

  • Spawn the review subagent to perform standard and spec audits on the branch's git diff.
  • The review agent outputs findings under ## Standards and ## Spec headers.
  • Evaluate results:
    • If there are findings (e.g., hard violations or baseline smells):
      • Send a message to the development subagent containing the reviewer's report.
      • Instruct the developer to resolve the bugs or smells.
      • Rerun the review agent once the developer has updated the branch.
    • If there are 0 findings:
      • Check tasks/todo.md for the next incomplete task.
      • If tasks remain, loop back to Step 2.
      • If all tasks are completed, proceed to Step 4.

4. Publishing Phase (Deployment Gate)

  • Spawn the publisher subagent to automate the integration and deployment lifecycle.
  • The publisher agent must:
    • Verify git status and stage all changed files.
    • Commit files using the Conventional Commits syntax and push to origin.
    • Execute gh pr create with standard metadata titles/descriptions to open the GitHub Pull Request.
    • Return the direct URL link to the Pull Request.

Exit Criteria

  • All tasks in tasks/todo.md are marked [x].
  • The review agent reports 0 standard violations or spec omissions.
  • All unit and integration tests in the test suite pass successfully.
  • The publisher agent has successfully pushed and opened a GitHub Pull Request, sharing its URL.

Sub-Agent Skill References

To ensure execution consistency, each sub-agent in this loop is dedicated to executing a specific workspace skill. They must load and re-use the following guides:

  • Planning Sub-Agent:
    • Primary Skill: planning-and-task-breakdown (used to analyze dependencies, partition slices, and output plans).
  • Development Sub-Agent:
    • Primary Skill: test-driven-development (drives test-first RED-GREEN loops at public boundaries).
  • Review Sub-Agent:
    • Primary Skill: code-review (performs two-axis checks for standard compliance and specification compliance).
  • Publisher Sub-Agent:
    • Primary Skill: pull-request (automates git checkout, commits, pushes, and GitHub CLI PR creation).

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.

Reviews can't be read right now.

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.