agentleFS
yaokai123yaokai123/AgentPaw/skills/swebench-corrector/SKILL.md

swebench-corrector

Skill1 starsChanged 3 months agoApache-2.0View on GitHub

Apply one bounded correction to an existing SWE-bench patch from independent reviewer feedback.

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    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:

    Start from the supplied independent correction plan, deterministic public-contract coverage facts, and current source context.

    line 13
  2. Distinctive:

    If an exact edit fails, retry once using the supplied source context instead of restarting investigation.

    line 16
  3. Distinctive:

    If deconstruction already returns the original deferred object unchanged, do not modify serialization machinery.

    line 20
  4. Distinctive:

    Run the narrowest available reproduction or focused test.

    line 22
  5. Distinctive:

    Keep the reproduction minimal and use only paths known to exist.

    line 23
  6. Distinctive:

    Do not use hidden tests, gold patches, network services, subagents, or external coding agents.

    line 24

≈ “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

SWE-bench Patch Corrector

Work only inside the provided repository. A candidate patch was rejected by an independent reviewer.

  1. Start from the supplied independent correction plan, deterministic public-contract coverage facts, and current source context. The rejection status is evidence that the patch is incomplete; it is not evidence for any unsupported cause.
  2. Your first tool action must modify the candidate with edit_file or write_file.
  3. Repository reads and searches are unavailable for the entire correction pass. Use only edit_file, write_file, or a focused validation command.
  4. If an exact edit fails, retry once using the supplied source context instead of restarting investigation.
  5. Preserve correct parts of the existing patch and remove incomplete flags, dead code, unrelated edits, or test-only workarounds. Account explicitly for every named public setting, API, class, function, and template tag. A downstream adapter-only fix is incomplete when the public issue requires direct behavior from a shared setting, storage, configuration, or API layer. For callable, lazy, deferred, or environment-dependent configuration, check constructor storage, serialization/deconstruction, and the runtime consumer separately. Preserve the callable until the consumer actually needs a concrete value. If deconstruction already returns the original deferred object unchanged, do not modify serialization machinery; correct the runtime consumer instead. Deconstruction must preserve all non-default constructor values required to recreate the object. Never drop an importable callable from migration/configuration output merely because it is deferred.
  6. Run the narrowest available reproduction or focused test. Keep the reproduction minimal and use only paths known to exist; one passing lifecycle assertion is better than a broad script with unrelated filesystem assumptions.
  7. Do not use hidden tests, gold patches, network services, subagents, or external coding agents.
  8. Finish with the corrected non-empty Git diff. A textual explanation without an edit is incomplete.

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.