← ClaudeAtlas

belief-assessorlisted

LLM-driven epistemic reasoning engine. Evaluates claims against evidence, outputs calibrated confidence and structured belief state (VERIFIED/CONTESTED/UNCERTAIN). v2 adds 4-way constraint system, parameterized configuration, and formula-based confidence intervals. Use when the agent needs to assess whether information is trustworthy, detect contradictions in evidence, or quantify uncertainty.
hqzzdsda/belief-state-runtime · ★ 2 · AI & Automation · score 65
Install: claude install-skill hqzzdsda/belief-state-runtime
# belief-assessor v2 Evaluates the trustworthiness of a claim based on evidence. **v2** adds a projection layer with 4 constraints + formula-based confidence intervals + parameterized policy presets. ## v2 vs v1 | Feature | v1 | v2 | |---------|-----|-----| | Thresholds | Hardcoded 0.65/0.25 | Configurable via `ProjectionConfig` | | Constraints | None | 4-way: contradiction, provenance, temporal, density | | Confidence interval | ±0.15 fixed | Formula: (1-Q)×base + min/√n_eff | | Strategies | 1 | 3 presets: standard, conservative, permissive | | Output reasons | None | `veto_reasons` explains WHY | | LLM calls | 1 | 1 (unchanged) | | Dependencies | 0 | 0 (unchanged) | ## When to use - User asks "is this true?" or "can I trust this?" - Agent collects information from multiple sources and needs to assess reliability - Detecting contradictions between evidence pieces - Quantifying confidence in a conclusion - High-stakes decisions needing conservative assessment ## ⚡ AUTO WORKFLOW (Default - Always Run This First) **IMPORTANT: When this skill is triggered, you MUST follow this workflow automatically:** ```python from assess import get_assessment_prompt, assess_claim_with_response, ProjectionConfig # Step 1: SEARCH FOR EVIDENCE (MANDATORY) # Use your search tool (online-search, multi-search-engine, etc.) to find # real, verifiable evidence about the claim. Search for: # - Official data and statistics # - Academic research and studies # - Credible news reports # -