agentleFS
okgritokgrit/done-is-a-claim/skills/collecting-worker-results/SKILL.md

collecting-worker-results

Skill33 starsChanged 37 days agoApache-2.0View on GitHub

Inspect a stopped worker's returned work and evidence before accepting it, retrying it or handing it to an integration owner. Use when a worker finishes, records disagree, or a green result has no visible change. Collection does not authorize merging.

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

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the report and blockers, then inspect the returned base, tip, commits, tracked diff and untracked files.

    line 27
  2. Distinctive:

    Read the tested revision, input hashes, command, environment and actual result before its success flag.

    line 33
  3. Distinctive:

    Before retrying a partial or ambiguous operation, inspect existing output, current state and any successor task.

    line 44

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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

Collecting worker results

A worker stopping, delivering a change and satisfying a task are different events. Reconcile the report, returned files and verification before deciding what was delivered. These are operational checks, not incident statistics.

Separate the records

ObservationSupportsDoes not establish
Process exit or runner statusThe recorded process outcomeA delivered change or acceptance
Worker reportThe worker's account of work and blockersIndependent verification
Commits, patches and working-tree diffA change footprint against a named baseCorrectness or integration
Acceptance recordThe result on its recorded input treeA pass on a different returned tree
Destination stateWhat reached the target revision or locationThat it matches the tested work

Procedure

  1. Confirm the run identity and that the worker stopped; a stale status or another process with the same name is insufficient. Collect only the assigned result. Resolve competing ownership before changing shared state.
  2. Read the report and blockers, then inspect the returned base, tip, commits, tracked diff and untracked files. A clean tree with no new commits may mean no change, an already-delivered change or a patch elsewhere; investigate.
  3. Compare the footprint with the task scope. A report-only return may satisfy a reporting task, but is not an implementation. Identify unrelated or stacked changes without assuming commit dates reliably establish ownership.
  4. Read the tested revision, input hashes, command, environment and actual result before its success flag. If the check measured an unchanged baseline, it demonstrates baseline behavior, not a requested implementation. An unchanged return can still satisfy an explicitly observational task.
  5. Rerun the agreed acceptance on the returned input tree yourself, plus the relevant regression checks. Preserve the command's exit status and log. A changed or incompletely identified tree needs new evidence. If checks cannot run, report that limit; do not translate it into a pass.
  6. Reconcile discrepancies: a failed packaging step can coexist with tested work, and a successful process can deliver none. Investigate environment and baseline failures rather than assigning cause from a label alone.
  7. Before retrying a partial or ambiguous operation, inspect existing output, current state and any successor task. Record the cause and changed brief; an earlier nonzero exit does not prove nothing was created.

Handover

Return task identity, base, delivered branch or patch, tested tree and hashes, commands/results, evidence locations, blockers and the proposed next step. State whether the work is only returned, verified or already integrated. Hand verified work to its integration owner; do not automatically commit, merge, cherry-pick or close another owner's task during collection.

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.