agentleFS
boaz-hwangboaz-hwang/distill/skills/refine/SKILL.md

refine

Skill14 starsChanged 36 days agoMITView on GitHub

Refine one existing skill's overall structure and guidance while preserving its intended outcome. Use when a skill is verbose, redundant, overprescriptive, unclear, or needs to be redefined around essential requirements and model autonomy.

42 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use addskill, apply-firmware and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • ryanportfolioSkill

    addskill

    Create, import, update, or install repo or personal skills: runtime ownership, resources, discovery checks.

    Also in 6 projects that use this

  • ryanportfolioSkill

    apply-firmware

    Bring current folder (plain folder, existing repo, old or partial firmware copy) up to full Harness Firmware in place. Mirror external repo → adopt-repo.

    Also in 6 projects that use this

  • +10 more from ryanportfolio

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 8 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 36 edited versions we read keep these.

  • 8

    Distinctive

    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 target's instructions and linked templates, references, and scripts.

    line 15
  2. Distinctive:

    When intent or conflicting requirements would change the rewrite, propose an interpretation and ask a focused question.

    line 20
  3. Distinctive:

    Keep the existing name unless renaming is requested or agreed.

    line 23
  4. Distinctive:

    Keep a recoverable baseline and prepare the revision separately.

    line 27
  5. Distinctive:

    Check the revision against the original purpose, essential requirements, metadata, and referenced resources.

    line 46
  6. Distinctive:

    Use comparable task conditions if claiming better execution quality or efficiency.

    line 55
  7. Distinctive:

    Do not weaken acceptance criteria to make the revision pass.

    line 56
  8. Distinctive:

    If a material requirement or necessary quality judgment remains unresolved, retain the candidate separately and ask for the specific …

    line 62

“Common” is measured: kept by at least 6 in 10 of the 36 edited versions of this file we could read. “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

Refine

Make one skill clear and sufficient for a capable agent. Review deletion first; preserve what quality depends on. Whole-skill restructuring is in scope, combining multiple skills is not. Discuss in the user's language.

1. Establish the intended result

Read the target's instructions and linked templates, references, and scripts. Infer its intended output, essential constraints, and quality criteria from the skill, available accepted examples, and user feedback. Execution records can help but are not required. Do not confuse today's inputs with reusable requirements.

When intent or conflicting requirements would change the rewrite, propose an interpretation and ask a focused question. Ask before changing the purpose or essential quality criteria unless the user has already authorized that change. Keep the existing name unless renaming is requested or agreed.

2. Refine the whole skill

Keep a recoverable baseline and prepare the revision separately. Find the shortest sufficient process for the intended result. Review deletion, merging/shortening, and revision before adding instructions. Remove repeated or irrelevant guidance and implementation detail a capable agent can decide; preserve non-obvious domain knowledge, fragile operations, and essential exceptions.

For each step, retain necessary inputs, decisive constraints, and what completion looks like. Keep required user decisions and permissions, without inventing extra approval gates. Reorganize the whole structure when it improves clarity; do not rewrite sound instructions merely for stylistic uniformity.

Inspect resources before removing or changing them. Bundle only what the revised skill needs, using relative paths; do not hide excess prose in supporting files. Reuse or derive needed templates from accepted artifacts. Ask for a missing resource only when it cannot be faithfully reconstructed. Make incidental values inputs. Choose languages and tools for the task, without adding unnecessary dependencies.

3. Check what changed

Check the revision against the original purpose, essential requirements, metadata, and referenced resources. Retain meaningful checks and their invocation in the skill; add missing evaluation only where it measures actual quality. Instruct the skill to run checks before delivery, report failed or untested criteria, and present concrete review questions only for judgments code cannot establish.

Match verification to the impact of the edit. Clear wording or duplicate removal may need only a requirements comparison. Changes to behavior, templates, or code need relevant checks and representative examples, including rare affected cases. Use comparable task conditions if claiming better execution quality or efficiency. Do not weaken acceptance criteria to make the revision pass. Distinguish structural cleanup from measured behavioral improvement; agent judgment is not human approval.

4. Deliver the revision

Apply a supported revision without overwriting unrelated edits. If a material requirement or necessary quality judgment remains unresolved, retain the candidate separately and ask for the specific missing decision. If nothing useful needs changing, keep the original.

Report the significant deletions or revisions, what was verified, and any remaining human 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.

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.