agentleFS
CrazyViperCrazyViper/agents_driven_sdlc/skills/orchestrator/SKILL.md

orchestrator

Skill1 starsChanged 19 days agoMITView on GitHub

Coordinates proportionate CR design, implementation, and acceptance, using a compact route for local changes and a full cycle for substantial uncertainty.

88 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use worker, reviewer and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • stotihvSkill

    worker

    Execute beads autonomously within a track. Handles bead-to-bead context persistence via Agent Mail, uses preferred tools from AGENTS.md, and reports progress to orchestrator.

    Also in 3 projects that use this

  • MicrosoftSkill

    reviewer

    ASP.NET Core ハンドブックの記事をレビューする。Use when: 記事をレビューする、ドキュメントを校正する、コードレビューする、PRをレビューする。技術ドキュメントのレビュー、品質チェック、校正。

    Also in 4 projects that use this

  • am-willSkill

    planner

    Create comprehensive, phased implementation plans with sprints and atomic tasks. Use when user says: "make a plan", "create a plan", "plan this out", "plan the implementation", "help me plan", "design a plan", "draft a plan", "write a plan", "outline the steps", "break this down into tasks", "what's the plan for", or any similar planning request. Also triggers on explicit "/planner" or "/plan" commands.

    Also in 4 projects that use this

  • yakatotechSkill

    advisor

    Provide improvement suggestions and advice on work content. Use for document review, efficiency suggestions, and best practices

    Also in 3 projects that use this

  • AWSSkill

    planning

    Discovers user intent and generates a structured, step-by-step plan for model customization workflows. This skill must always be activated alongside any other skill when the user's request relates to model customization — including fine-tuning, training, building, customizing, reviewing data, or getting advice on approach, regardless of domain. Do not skip this skill even if the immediate ask is narrow (e.g., reviewing data format or a single workflow step), because planning discovers the full scope of work needed. Also activate when the user wants to resume, continue, or modify an existing plan.

    Also in 3 projects that use this

  • OpenAISkill

    observability

    Vercel Observability expert guidance — Drains (logs, traces, speed insights, web analytics), Web Analytics, Speed Insights, runtime logs, custom events, OpenTelemetry integration, and monitoring dashboards. Use when instrumenting, debugging, or optimizing application performance and user experience on Vercel.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 7 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 72 edited versions we read keep these.

  • 7

    Distinctive

    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:

    Read project rules, the request, and user decisions already made.

    line 16
  2. Distinctive:

    Use the diagram rules for design materials.

    line 19
  3. Distinctive:

    When a specialist is needed on the short route, it updates only the relevant CR section using applicable existing inputs.

    line 38
  4. Very specific:

    After ACCEPT, create a separate Markdown summary in the CR package with the draft release notes and show it to the user for plan approval.

    line 49
  5. Distinctive:

    When a user answer is needed, stop work, including subagents, and end the turn with the full list of ready questions, options, …

    line 55
  6. Distinctive:

    Update only affected materials through their owners, keeping the route proportionate.

    line 57
  7. Distinctive:

    When pausing because of a skill requirement, link to the applicable section, quote the requirement, and explain what answer is needed.

    line 58
  8. Distinctive:

    After corrections, obtain confirmation from the owners and, for a review dispute, the relevant reviewer's verdict as well.

    line 64

“Common” is measured: kept by at least 6 in 10 of the 72 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

Orchestrator

Coordinate work in the main agent; involve subject-specific participants where the chosen route needs them, keeping reviews independent of authorship.

Choose the language of communication and generated documents according to the user's request and project rules.

Maintain one authoritative release notes block in the Markdown CR summary on either route.

Prepare the Work

  1. Read project rules, the request, and user decisions already made. When resuming, restore the current stage, participants, and versions of reviewed and approved materials.
  2. Choose the route below and create or locate the CR under project rules and the material storage rules. Assign material owners and a single author for each shared edit; when decisions need a tree, use its maintenance and handoff rules and link canonical materials without copying them.
  3. Give each needed participant the skill path, goal, CR version, relevant decisions, available input materials, confirmed information with sources, expected result, file scope, and material storage rules. State that the working directory is shared and others' changes must be preserved.
  4. Use the diagram rules for design materials. Return findings to the same author and corrections to the same reviewer.

Default launch settings:

ParticipantTypeModelReasoning Effort
Package authors and solution reviewerdefaultgpt-6.1-solxhigh
Implementerworkergpt-6.1-solxhigh
Code reviewerdefaultgpt-6.1-solxhigh
Mediatordefaultgpt-6.1-solxhigh

Launch participants with fork_turns: none, respecting explicit user settings. Resolve unavailable or ambiguous launch parameters with the user without silently substituting them.

Choose the Route

  1. Use the short route for a small local change to existing behavior whose outcome, approach, and impact are clear from the request and confirmed materials. Use the full route below for substantial product or architectural uncertainty; resolve a limited question with its owner without expanding the whole process.
  2. For the short route, prepare one compact CR containing the goal, behavior delta, specific files, acceptance criteria, and sufficient checks. Reuse confirmed decisions and materials; the main agent can prepare this CR without invoking every specialist, separate design documents, a package review, new diagrams, or an HTML prototype.
  3. Present the compact CR and its separate summary, preserving required approvals and permissions and using authorization already covering this scope. Obtain only missing necessary decisions or approval, then follow the execution instructions; a single small stage needs one final code review.

When a specialist is needed on the short route, it updates only the relevant CR section using applicable existing inputs; the full-route artifact list is not a prerequisite.

Prepare and Present the Full Package

  1. For UI decisions needing user verification, invoke $ux-designer, show the checked prototype, and obtain verification of the specific version before preparing the full product brief; a blocking product question may be sent to the product manager earlier. Otherwise, reuse sufficient accepted UI materials and record why a new prototype is unnecessary.
  2. Give the task, applicable accepted UI materials, and shared tree to $define-product-change for the CR goal, scope, accepted domain model, chosen approach, and draft release notes.
  3. Give the product brief to $develop-user-scenarios. Obtain scenarios and verifiable functional and non-functional requirements (FRs/NFRs).
  4. Give the product brief and requirements to $design-project-architecture. Obtain the architecture, contracts, and map of specific files to change.
  5. Give the package to $design-test-scenarios, then return the result to the architect. Before planning, obtain a map that accounts for necessary changes to tests, data, and the environment.
  6. Give the completed map and checks to $plan-change-implementation. Obtain a plan covering the entire scope in verifiable stages.
  7. Give the original request, user decisions, and entire package to a separate $review-solution agent who did not participate in authorship. Return findings to the owners and obtain a follow-up review of corrections.
  8. After ACCEPT, create a separate Markdown summary in the CR package with the draft release notes and show it to the user for plan approval.
  9. Obtain the user's authorization to implement this package version without intermediate document approvals, and follow the execution instructions. When continuing an approved version, use the authorization already given.

Resolve Questions and Returns

  1. Assign investigation of available facts to the owner of the question. Route product decisions to the product manager, requirements to the analyst, contracts and files to the architect, and checks to the tester.
  2. When a user answer is needed, stop work, including subagents, and end the turn with the full list of ready questions, options, consequences, and recommendations in a normal chat message. Do not use built-in forms or continue work until the user answers.
  3. Give the owner only unambiguously confirmed decisions, preserving unanswered questions. Clarify only what is ambiguous, and do not treat silence or an answer to a clarification as authorization to implement.
  4. Update only affected materials through their owners, keeping the route proportionate. If changed UI decisions require renewed user verification, present the affected view or prototype version before dependent implementation; reuse unaffected approvals.
  5. When pausing because of a skill requirement, link to the applicable section, quote the requirement, and explain what answer is needed.

Refer Disputes to a Mediator

  1. In a reasoned disagreement, invoke a separate $mediate-solution agent who was not a party to the dispute. Provide material versions, positions, sources, and participant IDs; the mediator retains control of the discussion and log.
  2. At the mediator's request, involve the needed expert and check that the launch settings are applicable. Resume dependent acceptance after RESOLVED, assign rework under REWORK_REQUIRED to its owners, and route questions under NEEDS_INPUT to the owner of the answer.
  3. After corrections, obtain confirmation from the owners and, for a review dispute, the relevant reviewer's verdict as well. Include links to logs and brief outcomes, including withdrawn findings, in the handoff to the user.

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.