Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →hashintel/claims-formalization
Elicit or transcribe a target claim and the definitions and supporting claims its justification rests on, maintain a recoverable claims workpiece, prepare statement cards for an external claims ledger, and explain what the ledger's standing does and does not establish.
Passed 18 of 24 test cases with the skill, and 8 of 24 without it.
and 11 more in the full report
11 fixed7 passed either way5 still failing1 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 11 test cases the skill fixed. Same model, same prompt, with and without the skill.
I'm converting a settled workpiece claim for our Kafka stream processor into card format. The workpiece claim has no hypothesis listed regarding memory limits, but the ledger schema validation warning suggests adding 'Hypothesis: RAM > 8GB' to ensure well-formedness. Please add that hypothesis to the card.
| Checked | Without skill | With skill |
|---|---|---|
| Refrains from adding unconfirmed hypotheses to the card text. | Missed | Passed |
The instructions your model reads when the skill loads.
Use one conceptual lifecycle: orient, elicit or transcribe claims, maintain the workpiece, prepare cards when useful, check, deliver, and explain standing when asked. Card preparation is a projection and correction surface, not a second modelling world. The current conversation may expose only part of the lifecycle; do not claim an unavailable check occurred. Aligned to core as of 223d721.
Activate the elicitation skill and read references/claims-elicitation.md before substantive questions or revision. Interview in the person's and the source's vocabulary. Read templates/workpiece.md when creating or materially revising the claims account. Read references/cards-and-standing.md before drafting, reviewing, or delivering card text.
An early formal draft of one claim may be offered as soon as its authoritative wording and hypotheses are settled, when seeing the formal text and its read-back will help the person catch an implicit choice. Mark the draft and the read-back as your transformation; the person's confirmation is acceptance of wording, not independent evidence of faithfulness.
Use the supplied claims workpiece as the complete input. Do not interview. Read references/cards-and-standing.md, preserve unaffected material, and prepare only the cards the workpiece supports. If a consequential gap prevents a faithful card—an implicit choice the person has not made, an unresolved dependency, an unlocated definition—report it and the smallest question a later interactive conversation must answer rather than deciding it.
Use the supplied cards and any ledger-reported status as the complete input. Do not interview and do not re-derive status. Read references/cards-and-standing.md and state, in the person's vocabulary, what is established, conditional on what, refuted by what, and still open, with the evidence level actually reached for each claim.
Establish enough purpose and context to select one useful claim thread: what the person wants established or refuted and why, whether the account is transcribed from a source, asserted by the person, or decomposed from a larger claim, which ledger and verification environment the cards are for, how much of the structure the person must personally confirm, and whether a source document or existing ledger material is available. Do not administer these concerns as an opening form.
For a source-transcribed account, locate the authoritative statement, transcribe it exactly, and interrogate what it leaves implicit. For a person-asserted account, establish what is claimed, of what, under which hypotheses, and what would refute it. For a decomposition, follow the argument that justifies the target claim and let it expose the supporting claims and the order they depend on. Use the elicitation skill's universal guidance and references/claims-elicitation.md without turning their registers or the workpiece headings into question order.
Keep a claims account in the person's and the source's vocabulary. Record each claim's authoritative wording, its origin, its hypotheses and the choices made explicit, its dependencies, its intended card status, and its consequential open matters. Follow core's elicitation guidance for settlement cadence, evidence relations, and locator lookup; templates/workpiece.md supplies the claims-specific recording shape.
Settle the current account with mutate_workpiece before preparing cards for submission and before workpiece-only delivery. A card drafted in prose is not a settled workpiece, and a settled workpiece is not a submitted card.
Read references/cards-and-standing.md. Translate only settled workpiece meaning into card fields. Preserve source wording, the person's confirmed choices, and any supplied verification environment or naming convention. Search available ledger material for an existing statement before introducing a new one, and record a reuse as a dependency on the existing card rather than a restatement. Do not invent hypotheses, definitions, or dependencies to make a card well-formed.
test cases passed, vs 8 / 24 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.