agentsclimarketplace

Coin the missing concept

Skill XyraSinclair/ideonomy/skills/coin-the-missing-concept

Computational ideonomy: Gunkel's science of ideas as inference-time machinery — a 37-primitive organon, an MDL-ratcheted respiratory engine, and 14 gated agent skills (Claude Code plugin)

Install
npx -y skills add XyraSinclair/ideonomy --skill coin-the-missing-concept

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

  • 0 stars0 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

When the same awkward phrase keeps recurring — "the kind of bug where...", "this class of lines that...", "that pattern we keep tripping on but have no name for" — treat the repetition as residue signaling a missing concept. Coin a precise noun, give it a decidable membership test, and keep it only if the corpus compresses. The failure it prevents: endless circumlocution standing in for a concept, where the model repeatedly gestures at the same structure but never binds a handle that lets the structure be reused or tested. Triggers on repeated locutions, taxonomic awkwardness, and any review where the same explanation is written three times with different wording.

SKILL.md

3.1 KB, as published. Nobody here has run it

coin-the-missing-concept

Grothendieck's concept-making fused with Feynman's bit-measuring and von Neumann's invariant discipline: a new noun earns its keep only by compressing the structure and predicting more than the old prose could. Organon: P8 name + P36 distill + P7 typify + P37 residue-seed. Phase: ORIENT. Respiratory: the MDL ratchet keeps the name only if the breath gets shorter and sharper.

Procedure

  1. Collect the repeated locutions. Gather the awkward phrases, examples, and near-duplicates that seem to point at one unnamed structure.

  2. Isolate the type behind them. Separate what is essential from what is incidental until the repeated cases read as members of one candidate type.

  3. Coin the smallest precise name and definition. Write one noun phrase and a one-sentence definition that says what belongs and what does not.

  4. Make membership decidable. Give a test another reader could apply to a fresh case without seeing the original examples.

  5. Measure compression, then predict outward. Re-express the corpus with the new concept and ask whether total description length fell. Then use the definition to identify a real case that was not on the original list.

  6. Keep or kill the concept. If it compresses and predicts, keep it. If it merely sounds elegant, delete it.

The gate

After coining, L(corpus|structure) must drop sharply, L(structure) must not rise enough to erase the gain, and the definition must predict a real previously-unlisted case. If any of those fail, the concept is decorative and dies. A clever label with no compression benefit is not a concept; it is copy.

This is a soft gate — for prose corpora the MDL terms are judged, not computed (see the gate-hardness classification in PREMIER_SKILLS.md). The out-of-sample prediction clause is the hard part of the gate; lean on it.

Example

A review keeps repeating "changes that fix the visible page but leave the coverage claim local." Coin the type: denominator-local fix — a fix that resolves a seen instance without stating or covering the full case space. Membership test: does the change land one example while the denominator remains unstated? Applied back to the notes, three clumsy paragraphs compress to the one term, and it correctly flags an untouched corpus-ingestion patch as another member. The noun earns its keep.

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.