agentleFS
jaydubya818jaydubya818/MissionControl/skills/workspace-handoff-checklist/SKILL.md

workspace-handoff-checklist

Skill2 starsChanged 8 months agoMITView on GitHub

Use this skill when a Mission Control work item is complete or awaiting review and an operator needs a factual handoff draft with outcome, evidence, unresolved risk, next owner, and next action.

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    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. Very specific:

    When an input is absent, write Needs confirmation for that field.

    line 28

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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

Workspace Handoff Checklist

Purpose

Produce a concise, factual operator handoff for a completed or review-ready Mission Control work item. Preserve uncertainty rather than filling gaps with assumptions.

Required inputs

Collect the following before drafting:

  1. Work-item identifier and current state.
  2. Confirmed outcome or current decision.
  3. Evidence references, such as a task, run, test output, document, or PR.
  4. Unresolved risk, blocker, or explicit confirmation that none is known.
  5. Named next owner and the single next action.

When an input is absent, write Needs confirmation for that field. Do not infer a status, URL, owner, approval, or completed action.

Workflow

  1. Confirm whether the work item is complete, awaiting review, or otherwise active.
  2. Extract only facts supported by the provided work-item context and evidence.
  3. Identify gaps, contradictions, or unverified claims and label them Needs confirmation.
  4. Draft the handoff using the required format.
  5. Self-check that all five required sections are present and that the draft does not change task state, request approval, send a message, create a commit, or deploy anything.

Required output format

## Handoff — <work-item identifier>

**State:** <complete | awaiting review | active | Needs confirmation>

**Outcome:** <confirmed result or Needs confirmation>

**Evidence:**
- <verified reference>

**Unresolved risk:** <known risk, blocker, or No known unresolved risk>

**Next owner:** <named person or role, or Needs confirmation>

**Next action:** <one explicit action, or Needs confirmation>

Examples

Complete evidence

Input: SFRL-073 is awaiting review. The local lint passed at 92/100. Draft import created version 0.1.0. Product Owner must decide whether to publish.

Output:

## Handoff — SFRL-073

**State:** awaiting review

**Outcome:** The candidate skill passed local structural lint and was imported as draft version 0.1.0.

**Evidence:**
- Local `skill-lint` result: 92/100
- Registry draft: `software-factory/workspace-handoff-checklist` v0.1.0

**Unresolved risk:** Registry evaluation is a structural proxy until an external agent runner is connected.

**Next owner:** Product Owner

**Next action:** Review the evaluation evidence and decide whether to publish the draft.

Missing evidence

Input: The task may be done, but no test output or owner was supplied.

Output:

## Handoff — Needs confirmation

**State:** Needs confirmation

**Outcome:** Needs confirmation

**Evidence:**
- Needs confirmation: no test output or source reference was provided.

**Unresolved risk:** Completion cannot be verified from the supplied context.

**Next owner:** Needs confirmation

**Next action:** Attach the relevant task and test evidence before review.

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.