agentsclimarketplace

Deliverable hygiene

Skill batteryshark/skill-tap/skills/development/deliverable-hygiene

Keep or restore a codebase as a finished, human-readable work product by separating agent scratch from source, pruning excess documentation and tests, deleting obsolete compatibility paths, and validating that process residue does not leak into the deliverable. Use during implementation cleanup or when a repository has visibly drifted.From its SKILL.md

Install
npx -y skills add batteryshark/skill-tap --skill deliverable-hygiene

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.8 KB, 513 tokens by cl100k_base, as published. Nobody here has run it

Deliverable hygiene

Keep the repository shaped like the artifact a human maintainer should inherit. Optimize for a clear mental model, a small useful surface, behavior-focused tests, and complete removal of obsolete paths.

Choose a mode

Use doctrine mode while adding or changing code:

  1. Read enough of the repository to understand its current model.
  2. Keep temporary reasoning and agent state outside the deliverable.
  3. Apply the rules in references/doctrine.md while editing.
  4. Validate that the final repository is easier to understand than the starting point.

Use recovery loop mode when the repository has accumulated residue:

  1. Run bin/deliverable-hygiene <repo> to collect review candidates.
  2. Audit concrete violations using references/filter-rubric.md.
  3. Fix high-confidence findings without widening the requested scope.
  4. Validate from repository evidence.
  5. Repeat for at most three passes, then stop GREEN, BLOCKED, or CAPPED.

Read references/workflow.md before recovery loop mode. Use references/examples.md when a classification is unclear.

Core rules

  • Treat the repository as the deliverable. Keep plans, prompts, review dumps, and loop state in disposable scratch space.
  • Put recovery state under .agent-work/deliverable-hygiene/<run-id>/ and ensure .agent-work/ is ignored.
  • Prefer names, structure, and concise local comments over new documents.
  • Preserve compatibility only when a public contract, persisted data, explicit requirement, or repository evidence makes it real.
  • Keep tests that protect meaningful behavior. Rewrite or remove tests that freeze private structure.
  • Add comments for intent, constraints, hazards, and external quirks that code cannot express.

Delegation

The portable role prompts in agents/ separate evidence gathering, editing, and validation:

Run the roles sequentially when independent subagents are unavailable. The fixer must verify the auditor's evidence, and the validator must inspect the repository rather than trust the fixer's summary.

Report

State the mode, categories fixed, files changed, checks run, final verdict, and any decision still blocked. Delete successful scratch before reporting.

What ships with it: 9 files

18.4 KB alongside SKILL.md, 2 of them executable

agents/

bin/

references/

scripts/

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.