From claim-harness
Use to search the canonical claim corpus before authoring a new claim. Finds related claims, reveals open gaps, prevents duplicates, and gives the agent a reading list of worked examples.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claim-harness:explore-corpusThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
The Polymer Claims corpus is the body of machine-verifiable claims at `github.com/beldez01/polymer-claims`. Before authoring a new claim, find out what has already been said.
The Polymer Claims corpus is the body of machine-verifiable claims at github.com/beldez01/polymer-claims. Before authoring a new claim, find out what has already been said.
Typical agent flow:
search_corpus(filter=...) via the claim-ir MCP.depends_on instead of duplicating.relations.extends: [...].relations.contradicts: [...] — disputes are first-class.| Tool | Purpose | Input |
|---|---|---|
search_corpus | Free-text + filter search | {text, domain, subject_kind, outcome, evidence_class, topic, freshness_days} |
query_neighbors | Embed a draft claim and return k-nearest | {claim_draft, k} |
fetch_claim | Full IR JSON for one claim | {claim_id} |
check_contradictions | Return known claims that oppose the draft on the same subject | {claim_draft} |
When the agent fetches a claim, walk it panel-by-panel:
provenance_state? If canonical_db, it's reproducible today; if local_file, the author ran it locally.outcome.Good. You have a novel claim. Proceed to /author-claim. But cite the absence — external_assumptions can carry an entry like {statement: "No prior claim covers this subject at this data version", kind: "design_choice", confidence: 0.9}.
npx claudepluginhub beldez01/polymer-claims --plugin claim-harnessProvides CDSS development patterns for drug interaction checking, dose validation, clinical scoring (NEWS2, qSOFA), and alert classification integrated into EMR workflows.