agentleFS
my-chiefmindmy-chiefmind/ai-maestro/skills/tpm/SKILL.md

tpm

Skill6 starsChanged 49 days agoMITView on GitHub

Run delivery control for the current Maestro project, ticket, swarm, or messy worktree. Reconciles board and repository evidence into a proceed/block decision; does not implement, merge, deploy, or silently clean up work.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 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.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill to establish the delivery truth for the current project or a named ticket.

    line 8
  2. Very specific:

    Run the board-validate and orchestration-health skills.

    line 25
  3. Distinctive:

    Do not create separate tickets for minor findings or reminders.

    line 64
  4. Distinctive:

    Stop agent work for an external blocker.

    line 67
  5. Very specific:

    After the table, include Drift or risk: only when board, branch, worktree, QA, PR, or merge state disagrees.

    line 91
  6. Distinctive:

    Do not say “almost done” or infer completion from a branch or PR alone.

    line 94

≈ “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

Delivery TPM

Use this skill to establish the delivery truth for the current project or a named ticket. It is a control and recovery entry point: it turns real board, branch, PR, worktree, and gate evidence into one next action. It is not an implementation, QA, merge, deployment, or status-narration workflow.

Choose the scope

  • Named ticket — assess that ticket before it starts, resumes, or proceeds through a gate.
  • Current board — assess active work, blockers, drift, and the one next eligible ticket.
  • Swarm — assess pool readiness at startup/recovery or when capacity, lanes, or merges are contended. Use the swarm skill for continuous dispatch; this skill supplies the proceed/hold decision.
  • Messy worktree or branch — assess ownership and safe recovery. Preserve all work; do not delete, recreate, or dispatch replacement work while ownership is unresolved.

Gather evidence before deciding

  1. Run the board-validate and orchestration-health skills. Read the board with maestro ticket read --board {{BOARD}}/data.json --json (the files plus any machine-local status); use the project’s supported board CLI for any authorized mutation, never a hand edit.
  2. Inspect the relevant repository state: primary checkout status, active worktrees and their branches, commits relative to the default branch, open/merged pull requests, and current test/QA/delivery evidence. Record unknown facts as unknown, not as complete.
  3. For a ticket, confirm it is neither deferred, duplicated, already delivered, nor externally blocked; its dependencies are complete; its acceptance criteria are measurable; and it has one expected delivery PR and verification command.
  4. For swarm, run maestro swarm status --json and maestro lanes next --board {{BOARD}}/data.json --json. Do not start lanes here. A board error, unexplained claim, ownership conflict, or invalid lane schedule is a hold condition.

Decide the smallest safe outcome

Return exactly one outcome per affected ticket:

  • Proceed — the ticket is ready for its named next stage. State that stage.
  • Resume — an owned in-progress ticket has a coherent worktree/branch and a concrete next stage.
  • Finish landing — the change merged but its archive/evidence/board reconciliation remains.
  • Blocked — give the reproducible failed command, unmet dependency, unavailable authority, credential/environment, or ownership conflict; name who can remove it and the smallest next action.
  • Owner decision required — a deferred, inactive, or blocked ticket has live work. Preserve it and ask the owner to choose resume, archive, or cleanup.

For a healthy swarm, return one pool-wide Proceed and only list its startable lanes. Invoke this skill again only when there is drift, stale work, merge contention, a recovery event, or new evidence that changes readiness. Do not consume a TPM slot for every healthy lane.

Handoff and boundaries

  • A Proceed hands a ticket to the orchestrator or the swarm coordinator. For an explicit ticket checkpoint, use "agent_plan": ["tpm", "pe", "backend"] (substitute the real implementation role).
  • QA independently reviews implementation. Principal Delivery makes the final land/no-land decision. TPM never replaces either role.
  • Do not create separate tickets for minor findings or reminders. Keep in-scope repairs with the current ticket. Only a genuinely independent release boundary, owner, or shared prerequisite warrants a new ticket.
  • Stop agent work for an external blocker. Never create speculative code or documentation-only work around a missing approval, credential, deployment authority, or real-user test.

Report

Lead with a qualitative project judgment based only on the evidence gathered:

  • Healthy — the board matches repository state, active work has a coherent next stage, and no current blocker prevents the eligible delivery flow.
  • At risk — delivery can continue, but drift, an aging gate/branch, unresolved evidence, or constrained capacity needs a named next action.
  • Blocked — an external prerequisite, invalid schedule, ownership conflict, failed gate, or missing authority stops the affected delivery flow.

Then return this table. Use short bullet fragments inside cells when more than one fact is needed; write — when there is none. Never manufacture a row or a percentage merely to make the report look complete.

ScopeDone / evidenceMissingBlocker / ownerNext
Project judgment — <Healthy / At risk / Blocked><facts supporting the judgment><unmet delivery gates or unknowns><concrete issue and owner, or —><one highest-leverage action>
Active ticket — <ID, title, stage><completed acceptance criteria, revision, checks><specific remaining criterion or gate><exact blocker and owner, or —><one next stage/action>
Other affected ticket / lane — <ID><evidence, or —><what it needs><blocker and owner, or —><next action>

After the table, include Drift or risk: only when board, branch, worktree, QA, PR, or merge state disagrees. State the evidence and remediation; otherwise say Drift or risk: none.

Do not say “almost done” or infer completion from a branch or PR alone. A project judgment is not a release declaration: individual tickets are done only after their applicable acceptance, verification, QA, delivery, merge, and board-reconciliation gates complete.

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.