agentleFS
pahlevikunpahlevikun/levistack/skills/agent-authoring/convert-as-skill/SKILL.md

convert-as-skill

Skill0 starsChanged 19 days agoMITView on GitHub

Convert existing material into a skill, merge existing skills into one, or enhance one skill by learning from upstream skills, repos and pages. Use when asked to convert a rule, command or agent to a skill, save a conversation as a skill, turn a PDF or docs folder into a skill, merge two skills into one, or improve a skill from other sources (keep, replace and drop each method, rewrite in our own words, test it). Not for writing a skill from scratch (`create-skill`).

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Classify first: Run scripts/classify.mjs before opening a GUIDE.

    line 38
  2. Distinctive:

    Third-party material stays private: Do not commit a generated book skill, a clone of an upstream repo or a source map to this catalog.

    line 42
  3. Distinctive:

    Treat "verbatim" as a use case.

    line 74
  4. Distinctive:

    Default "convert this" with no files to a conversation.

    line 75
  5. Very specific:

    Run convert.mjs on a PDF or a skill folder, or extract_document.py on rules/ / commands/ / agents/.

    line 76
  6. Very specific:

    Use enhance-skill to combine two skills (that is merge), or to turn a book into a skill (that is from-document).

    line 77
  7. Distinctive:

    Delete sources, or commit a third-party book dump to this catalog.

    line 79

≈ “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

Convert as skill

One skill, six use cases. A classifier picks exactly one, or asks. Load only that use case's GUIDE. Copy, extract, distill and compose are different jobs.

For a new skill with no source file use create-skill. For a new rule, command or agent use create-rule, create-command or create-agent.

Principles

  1. Classify first. Run scripts/classify.mjs before opening a GUIDE. File layout beats wording. Never default to conversation.
  2. One use case per input. A PDF is never copied as a rule. Mixed kinds are sequential jobs, or ask — not one blended skill.
  3. Copy vs extract vs distill vs compose. Rules and agents copy the body. A conversation extracts what worked. A document distills structure. Merge composes a new skill and leaves the sources. Enhance adds to one skill from other sources, in our own words.
  4. Preview, then write. Never delete on your own. Dry-run first. Remove originals only after the user has reviewed the result and asked.
  5. Third-party material stays private. Do not commit a generated book skill, a clone of an upstream repo or a source map to this catalog. They live in an ignored scratch folder.

Step 0 — classify

node scripts/classify.mjs [--text "<user request>"] [--json] [path...]
StatusDo
okState the kind and reason. Load only that GUIDE.
askAsk the printed question. Load no GUIDE.
rejectStop. Use the named create-* skill instead.

from-command uses the from-rule GUIDE; the kind stays distinct in logs.

Use cases

KindSourceWhat happensGUIDE
from-rule / from-command.mdc, rules/, commands/Copy the bodyfrom-rule
from-agentagents/*.mdKeep / wrap / convert, then copyfrom-agent
from-conversationChat / transcript, no filesExtract what workedfrom-conversation
from-documentPDF, docs folder, HTML, URLDistill structurefrom-document
merge-skillsTwo or more skill foldersCompose one new skillmerge-skills
enhance-skillOne skill folder plus upstream skills, repos, pages or a named conversationAdd to the skill: keep, replace or drop each method, write it ourselvesenhance-skill

Scripts: scripts/classify.mjs, scripts/convert.mjs (rules/agents only), scripts/extract_document.py, scripts/merge.mjs, scripts/enhance.mjs (read-only inventory and fetch plan). Document parsers live in scripts/extractor/ (pdf.py, html.py, docx.py, epub.py, text.py, rtf.py, sanitize.py, dependencies.py, config.py, exceptions.py, scan_generated_skill.py). Templates: harvest.md, merge-outline.md, enhance-map.md.

Do not

  • Open a GUIDE before classify, or load two GUIDEs for one input.
  • Treat "verbatim" as a use case. It is only how from-rule / from-agent copy the body.
  • Default "convert this" with no files to a conversation.
  • Run convert.mjs on a PDF or a skill folder, or extract_document.py on rules/ / commands/ / agents/.
  • Use enhance-skill to combine two skills (that is merge), or to turn a book into a skill (that is from-document).
  • Drop a useful source idea only because it was big. Replace it with a smaller version of your own.
  • Delete sources, or commit a third-party book dump to this catalog.

Related skills

  • create-skill: write a skill from scratch, or fix the result after conversion.
  • polyglot-copywriter: rewrite the harvested methods in simple, plain words.
  • second-opinion: have a second reviewer check an enhancement before you report it.
  • create-rule: the source is a rule and it should stay a rule.
  • create-command: the source is a saved prompt and it should stay a command.
  • create-agent: the source is a subagent and it should stay one.

Done when

  • Classify printed one kind (or asked / redirected) and only that GUIDE ran.
  • The new skill lints clean. Skipped or refused inputs have a stated reason.
  • The user knows which originals remain and how to remove or restore them.

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.