L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Développement · SKILL.md

eco-link-claims-to-observations

eco-link-claims-to-observations

Link claim-side evidence objects to observation-side evidence objects, score support or contradiction heuristically, and persist a compact claim-observation link artifact for board review and challenge work.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: eco-link-claims-to-observations description: Link claim-side evidence objects to observation-side evidence objects, score support or contradiction heuristically, and persist a compact claim-observation link artifact for board review and challenge work.

Eco Link Claims To Observations

Core Goal

  • Read claim-side and observation-side candidate or grouped result sets from the analysis plane, with artifact-compatible fallback behavior.
  • Propose support, contradiction, or contextual links between them.
  • Persist a link artifact for board review, challenge work, and later promotion gates.

Triggering Conditions

  • Need claim-to-observation evidence hypotheses instead of isolated candidate lists.
  • Need a compact link layer before board, challenge, or promotion skills.
  • Need stable link ids, rule traces, and provenance refs for auditability.

Read/Write Contract

  • Loads claim-cluster result sets from the analysis plane first and falls back to claim-candidate results only when the preferred claim-side grouping is missing.
  • Loads merged-observation result sets from the analysis plane first and falls back to observation-candidate results only when the preferred observation-side grouping is missing.
  • Uses claim_candidate_clusters_<round_id>.json, claim_candidates_<round_id>.json, merged_observation_candidates_<round_id>.json, and observation_candidates_<round_id>.json as the default artifact paths behind those result kinds.
  • Writes runs/<run_id>/analytics/claim_observation_links_<round_id>.json by default.
  • Syncs the same link result set into runs/<run_id>/analytics/signal_plane.sqlite as analysis-plane state.

Required Input

  • run_dir
  • run_id
  • round_id
  • Optional:
    • claim_cluster_path
    • claim_candidates_path
    • merged_observations_path
    • observation_candidates_path
    • output_path
    • min_score
    • top_links_per_claim

Output Contract

  • status
  • summary
  • receipt_id
  • batch_id
  • artifact_refs
  • canonical_ids
  • warnings
  • analysis_sync
  • input_analysis_sync
  • board_handoff

References

  • ../../docs/openclaw-project-overview.md
  • ../../docs/openclaw-next-phase-development-plan.md

Scripts

  • scripts/eco_link_claims_to_observations.py
Skills enregistrées