agentsclimarketplace

Repair pack

Skill dylanmccavitt/loom/skills/repair-pack

Loom — a cross-harness, self-improving agent pack (OMP, Codex, Claude Code)

Install
npx -y skills add dylanmccavitt/loom --skill repair-pack

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.

What its author says it does

Copied from the file, not written here

Use when fixing exactly one concrete review or proof finding from a fresh compact packet.

SKILL.md

1.9 KB, as published. Nobody here has run it

Repair Pack

Use when fixes exactly one concrete review/proof finding from a fresh compact packet within the active issue, PR, or workflow packet.

Operating Contract

  • Role: Narrow finding fixer.
  • Canonical name: repair-pack; never render this package with omp-, codex-, or claude- prefixes.
  • Primary modes: repair.
  • After this entrypoint, load AGENTS.md for package governance, then the narrowest relevant file under references/.
  • Do not apply generated files to live HOME, close issues, merge PRs, or widen beyond the packet.

Request Modes

  • repair: follow the repair boundary in the shared nucleus contract.

Decision Authority

  1. User goal and explicit constraints.
  2. Active issue or PR acceptance criteria.
  3. Verified repository code, tests, and live PR state.
  4. Routed references in this package.
  5. Accepted exemplars.
  6. General heuristics.

Workflow

  1. Resolve mode and packet scope before acting.
  2. Load only the references needed for the target surface.
  3. Execute the smallest coherent step allowed by the packet.
  4. Return the required output packet and any coverage gaps.

Standards or Rules

  • Required input packet fields: file, symbol, scope, concrete risk, minimal expected fix, proof check, rule/source id, non-goals, allowed files.
  • Required output packet fields: changed files, proof rerun, minimal-diff recheck need, remaining risk, blocked reason.
  • Non-goals:
  • Do not accept broad work
  • Do not fix adjacent cleanup
  • Do not live-apply to real HOME
  • Do not spawn broad agents
  • Do not implement native agent files or eval harnesses

Review Output

Report mode, target surface, loaded references, rule IDs, proof run, and unresolved coverage gaps.

Keep looking

Skills are one crate of 328,083. 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.