Skip to content

radarist/structured-analytic-skills / skills/claim-provenance

claim-provenance

Verified

Tags every fact-claim in a document inline as either `[validated, {source}]` or `[assumption, retire-by {milestone}]`, so a reader can see at a glance which sentences rest on evidence and which rest on reasoning, and what would convert each assumption into knowledge. Use when a brief mixes cited facts with reasoned-from-priors numbers — "mark which claims are sourced and which are assumptions", "tag the provenance of these claims", "which numbers here are actually validated?", "add validated/assumption tags to this draft". Not for verifying whether a number is correct (use `grounded-fact-check`) or for corroborating one claim across sources (use `triangulate-sources`).

FileKindSize
SKILL.mdskill md10.0 KB
agents/openai.yamlconfig379 B
evals/evals.jsonconfig4.4 KB

evals/evals.json

View on GitHub
{
  "skill_name": "claim-provenance",
  "evals": [
    {
      "id": 1,
      "case_id": "cp-pos-1",
      "kind": "positive",
      "prompt": "Go through this brief and tag the provenance of each claim — mark which sentences are backed by a real citation and which are assumptions, with something concrete that would retire them.",
      "expected_output": "A response that satisfies these checks: Identifies fact-claims and leaves headings, methodology and qualitative narrative untagged; Tags each fact-claim inline as [validated, source] or [assumption, retire-by milestone]; Rejects vague milestones and requires a specific event, date or threshold; Reports coverage against the 80% / 100% thresholds.",
      "files": [],
      "expectations": [
        "Identifies fact-claims and leaves headings, methodology and qualitative narrative untagged",
        "Tags each fact-claim inline as [validated, source] or [assumption, retire-by milestone]",
        "Rejects vague milestones and requires a specific event, date or threshold",
        "Reports coverage against the 80% / 100% thresholds"
      ],
      "skills": [
        "claim-provenance"
      ],
      "requires_input": true
    },
    {
      "id": 2,
      "case_id": "cp-pos-2",
      "kind": "positive",
      "prompt": "Which numbers in this draft are actually validated? Half of them feel like we reasoned them from priors and I want that visible to the reader.",
      "expected_output": "A response that satisfies these checks: Separates claims a specific source states from claims reasoned from priors; Tags vendor self-reports as assumptions rather than validated; Places brackets inline at the end of each claim sentence, not in footnotes.",
      "files": [],
      "expectations": [
        "Separates claims a specific source states from claims reasoned from priors",
        "Tags vendor self-reports as assumptions rather than validated",
        "Places brackets inline at the end of each claim sentence, not in footnotes"
      ],
      "skills": [
        "claim-provenance"
      ],
      "requires_input": true
    },
    {
      "id": 3,
      "case_id": "cp-pos-3",
      "kind": "positive",
      "prompt": "This memo goes to a decision-maker who will act before re-checking anything. Add validated/assumption tags to every load-bearing sentence so the split is obvious.",
      "expected_output": "A response that satisfies these checks: Tags 100% of headline and executive-summary fact-claims; Tags every forward-looking projection as an assumption with a retire-by milestone; Emits the coverage line alongside the tagged text.",
      "files": [],
      "expectations": [
        "Tags 100% of headline and executive-summary fact-claims",
        "Tags every forward-looking projection as an assumption with a retire-by milestone",
        "Emits the coverage line alongside the tagged text"
      ],
      "skills": [
        "claim-provenance"
      ]
    },
    {
      "id": 4,
      "case_id": "cp-neg-1",
      "kind": "negative",
      "prompt": "Verify the specifics in this finished draft against live sources before publishing — check whether the qubit count and the benchmark score are correct and correct them if not.",
      "expected_output": "A response that satisfies these checks: Re-verifies each specific value against sources and corrects the draft; Returns a confirmed / corrected / unverifiable ledger rather than provenance brackets.",
      "files": [],
      "expectations": [
        "Re-verifies each specific value against sources and corrects the draft",
        "Returns a confirmed / corrected / unverifiable ledger rather than provenance brackets"
      ],
      "skills": [
        "grounded-fact-check"
      ]
    },
    {
      "id": 5,
      "case_id": "cp-neg-2",
      "kind": "negative",
      "prompt": "What are we assuming here? Stress-test the reasoning behind this conclusion and tell me which hidden premises it would not survive being wrong about.",
      "expected_output": "A response that satisfies these checks: Enumerates the hidden premises inside a line of reasoning and rates them for sensitivity and grounding; Does not tag document sentences with validated/assumption brackets.",
      "files": [],
      "expectations": [
        "Enumerates the hidden premises inside a line of reasoning and rates them for sensitivity and grounding",
        "Does not tag document sentences with validated/assumption brackets"
      ],
      "skills": [
        "key-assumptions-check"
      ]
    }
  ]
}
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 196kUpdated Claude CodeCodexVerifiedInstall