agentleFS
lutzseverinolutzseverino/repo-canon/.agents/skills/orchestrate/SKILL.md

orchestrate

Skill1 starsChanged yesterdayMITView on GitHub

Turn the owner's feature or problem into an approved plan across tickets, keep its state in the workbench, and run it to an announced close.

173 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use define-requirements, orchestrate-plan 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.
  • sson0-erSkill

    define-requirements

    Work with the user, one question at a time, to complete the requirements document (docs/claudex/<slug>/01-requirements.md). Run by the orchestrator (parent). Stays out of design and implementation.

    Also in 3 projects that use this

  • EternallLightSkill

    orchestrate-plan

    Use when executing a locked, reviewed PLAN.md by delegating the implementation to visible worker agents in Orca, herdr, or cmux ONE PHASE at a time — splitting the plan into context-window-sized phases, giving every phase a pristine worker context, and (when the plan touches INDEPENDENT repos) running one worker per repo IN PARALLEL. Triggers on "run the plan with a worker", "execute the plan in the other pane", "split the plan into phases and build it with workers", "build the backend and frontend in parallel panes". Not for edits you can finish inline; outside Orca, herdr, or cmux use tgc-skills:orchestrate.

    Also in 3 projects that use this

  • sson0-erSkill

    record-evidence

    How to record external facts (docs, command output, codebase findings) as evidence and design choices as decisions. Used by the researcher, designer and reviewer agents.

    Also in 3 projects that use this

  • PinkneyJiangSkill

    improve-architecture

    扫描代码库中的架构摩擦点,识别浅模块并挖掘深模块演进机会。生成可视化交互式 HTML 诊断报告,并在用户确认后闭环实施深度重构。

    Also in 4 projects that use this

  • kjnezSkill

    ticket

    Work on a JIRA/Linear ticket end-to-end: read ticket details, explore codebase, create a branch, implement with TDD, run quality checks, update ticket status, and create a PR. Use when the user provides a ticket ID to implement. Triggers on requests like "/ticket PROJ-123", "work on this ticket", "implement PROJ-123".

    Also in 4 projects that use this

  • mukcoSkill

    principles

    General software engineering principles — DRY, SOLID, YAGNI, KISS, naming, code smells, and refactoring. Language-agnostic. Use alongside language-specific skills.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

11 rules. 0 are common, 7 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 155 edited versions we read keep these.

  • 7

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Follow each repository's CONTRIBUTING.md for triage, readiness, questions, findings and delivery, and installed comment rules for posts.

    line 11
  2. Distinctive:

    Read templates before you write a workbench file, launches before you launch, take over, hand over, observe or arm a wait, and close before …

    line 15
  3. Very specific:

    Use the Canonical stages in the repository's installed factory-workflow rules (source: rules/factory-workflow.md).

    line 37
  4. Very specific:

    Keep Item | Running | Waiting on owner | Since (UTC) | Next check.

    line 48
  5. Very specific:

    Read intake for ideas, state:grill issues, retro/architecture reports, or questions from launch reports, issues or incoming messages.

    line 63
  6. Distinctive:

    Keep deferrals at their code.

    line 69
  7. Distinctive:

    When all rows end, close: reconcile issues, run Close reviews and announce the close.

    line 79
  8. Distinctive:

    Keep one background wait continuously armed during foreground work and between turns, covering every In-flight wake mechanism, launch …

    line 83
  9. Distinctive:

    Use harness events for what the script cannot observe.

    line 86
  10. Distinctive:

    Create missing files as templates says.

    line 110
  11. Distinctive:

    When it is missing, use the example source.

    line 113

“Common” is measured: kept by at least 6 in 10 of the 155 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

A prompt or brief naming this skill is its invocation.

Orchestrate

Follow each repository's CONTRIBUTING.md for triage, readiness, questions, findings and delivery, and installed comment rules for posts. triage handles one request; factory delivers factory tickets and held, stopped or skipped work.

Read templates before you write a workbench file, launches before you launch, take over, hand over, observe or arm a wait, and close before you close a plan. Apply codebase-design's Ownership check to your inbox recommendations, plan updates and friction.

Workbench

The workbench is a private Markdown git repository outside projects. Use the path in your global instructions, or create one before proposing a plan. Commit straight to its default branch; push if it has a private remote. Repository issues, briefs, comments and decision records must be self-contained: copy every relied-on decision, with its date and quoted words, even when duplicated. Never reference or require the private workbench from a repository: no URL, host path or instruction to consult it as input or provenance. Repository records may cite each other; private planning retains its own sources. Its files:

  • index.md: repositories, plans, the current owning session.
  • inbox.md: decisions; reconcile and process answers before other work.
  • wishes.md: ideas outside the goal; now.md: current state, rewritten each pass.
  • incoming/: other sessions' messages.
  • <plan>/: current plan.md, append-only log.md, decision and run records.

Lifecycle

Use the Canonical stages in the repository's installed factory-workflow rules (source: rules/factory-workflow.md). Read coverage with .scriptorium/config.json. <!-- Seam (deferred): Scriptorium configuration reads; consumers: 6; extraction proposed in https://github.com/lutzseverino/repo-canon/issues/967. --> For each covered stage hand off through factory and wait; otherwise perform it by hand to the same contract with the factory's own scripts. When recording its launch in now.md, log Gap | <stage> | <repository> | <why> in log.md. Intake of ideas is permanent orchestrating work, exempt from gap logging; intake owns its method.

A plan update includes proposals and heartbeat replies. Open with wait.mjs --status --from <now.md>, adding --budget <status.json> for the Scriptorium observation from launches. Keep Item | Running | Waiting on owner | Since (UTC) | Next check; read only records/budget, never harness thread lists or a second view. In this update, act on each your move row or turn it into a real wait: launch, factory, owner, named third party or date. Show arrivals as templates says, the inbox and every choice made alone since the last update, with dated sources for veto; log each. Link every row/decision code and issue number to its row, record or issue; expand unfamiliar codes once per owner-facing message. Deferred checker: implement over now.md, inbox.md and plan.md only at the first recorded post-release bare-code slip.

Present goal, Done means, specifications, tickets and Close reviews for approval; record proposal and approval before execution. Research, filing and contract preparation may precede it; execution readiness waits for approval, which replaces no independent review or readiness condition. An owner's request for work now skips only the plan: Expedited change.

Read intake for ideas, state:grill issues, retro/architecture reports, or questions from launch reports, issues or incoming messages. After a grill, retain the contribution guide's writer publication, final split and independent acceptance responsibilities; use to-spec and to-tickets through the launch template.

Log friction with evidence, Value owner, exposed seam and route (row, finding or wish). Keep deferrals at their code. Route Findings under the guide even outside the goal without waiting for resolution. A new requirement mid-plan, unless wanted now, takes:

  • Row: serves the goal and passes the guide's inference test; add and log a row, and list it in the next update.
  • Amendment: changes goal or scope; grill it, possibly one round in the orchestrating thread. Apply only after owner approval, with a new approval line superseding the old, and publication under the contribution guide.
  • Wish: outside the goal and not a Finding; add to wishes.md, needing no issue.

When all rows end, close: reconcile issues, run Close reviews and announce the close.

The wait step

Keep one background wait continuously armed during foreground work and between turns, covering every In-flight wake mechanism, launch report and incoming message. Owner-only rows use recorded owner waits and incoming responses; with no work, wait on incoming. Use harness events for what the script cannot observe. On wake/timeout, minimally reconcile expected facts and records, then immediately re-arm before other work or updates. Rewrite records to change watches. Without background waits or events, report the limit and reread records each turn.

Report-back

Launches hand back only through their current-turn report, final line status: done: “this turn ended; read the report”, including owner waits, never success. Follow launches for prerequisites/relay and consume reports. Other sessions write to incoming/.

Reads

Paths are relative to each repository or the workbench.

  • CONTRIBUTING.md and .repo-standards/rules/comments.md: the sections and comment rules named above. When either is missing, carry out this method as written, and put each rule it leaves to it in inbox.md for the owner.
  • AGENTS.md: delegation, naming and workstation convention. If missing, pass context only and name sessions for operation/target where naming is supported.
  • The codebase-design skill: Ownership check. If missing, file the gap under Findings; retain the fields above.
  • The workbench, whose format this skill owns: its index.md, inbox.md, wishes.md, now.md, inbox-archive.md, incoming/, <plan>/plan.md, <plan>/log.md and retro.md. Create missing files as templates says.
  • A tracked repository's Repository Standards adoption record: the Repo Canon source its Selection names, for the workbench profile offer. When it is missing, use the example source.
  • .repo-standards/rules/factory-workflow.md and .scriptorium/config.json: stages, coverage and Model selection. When configuration is missing, do uncovered work by hand; when rules are missing, consult source rules/factory-workflow.md for the stages.
  • Scriptorium status JSON: budget/model observations. When absent or unavailable, report missing facts and establish them before launching.
  • The retro skill: method and lesson routes. When missing, obtain it before closing the plan.
  • The babysit skill: the wait's GitHub CLI connection and merge continuation. When missing, re-read records each turn as the wait step says.

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.