0 passed0 failed

Claim Discovery Workflow

Claim discovery is a high-recall proof-planning pass, not a verdict.

Map keys: promise.claim-discovery, rule.reviewable-artifacts, rule.evidence-gaps, rule.agent-human-resumability. Evidence path: deterministic plus skill review. Evidence status: open gap. Next action: keep source-scope tests, canonical-map generation, and review-result replay connected to the active spec tree. Terms covered here: source inventory, entry Markdown, linked Markdown, .gitignore, raw candidates, duplicate handling, canonical compression, review-result replay, false-positive review, false-negative boundary.

Maintainer Promise

discover claims emits source-ref-backed candidates from configured entry documents and the linked Markdown they reach, preferring recall and preserving the scan boundary so curation can distinguish binary false negatives from out-of-scope narrative gaps.

Subclaims

  • Discovery emits source-ref-backed candidates from configured entry documents and linked Markdown within the declared depth bounds.
  • Discovery favors recall; a missed declaration inside the scan boundary is a binary bug, while missing behavior outside the boundary is catalog, narrative, or alignment work.
  • Duplicate handling and false-positive curation remain packet-aware review responsibilities rather than deterministic verdicts.
  • The active scan boundary excludes archived spec trees and superseded claim pages so they do not dilute current proof planning.
  • Canonical compression and review-result replay consume the discovery packet without mutating discovery's recall behavior.

Evidence