Label AI claims with confidence levels and separate factual from symbolic output.
npx skills add https://github.com/TylerGarlick/abraxas --skill janus-system
Please help me install this Agent Skill. Skill: janus-system Source: https://github.com/TylerGarlick/abraxas/tree/main/skills/janus-system Command: npx skills add https://github.com/TylerGarlick/abraxas --skill janus-system
This Skill addresses the core problem of AI hallucination and output ambiguity by enforcing epistemic discipline, clearly labeling factual versus symbolic output, and providing transparency into the AI's reasoning process.
[KNOWN], [INFERRED], [UNCERTAIN], or [UNKNOWN] for factual claims, and [DREAM] for symbolic/creative content.Use the janus-system skill to get a factual answer about the alchemical process.
High-intent search queries and answers about installing and using this skill.
Preventing AI hallucination involves enforcing epistemic discipline by labeling factual claims with confidence levels like [KNOWN], [INFERRED], or [UNCERTAIN], separating objective facts from creative generation to ensure transparency.
Epistemic labeling enforces truth-discipline by tagging every factual claim with confidence levels and distinguishing symbolic content. It uses a two-face architecture with a threshold mechanism to prevent cross-contamination between factual and creative output streams.
Separating factual answers from creative generation involves using distinct output streams managed by a Threshold mechanism, preventing cross-contamination between factual and symbolic content during your AI session.
Tracking AI uncertainty across multiple chat sessions requires maintaining a cross-session Epistemic Ledger that persistently records findings, uncertainties, and anti-sycophancy events for ongoing accountability.
To inspect the internal state of an AI model, you can use a Qualia Bridge that allows querying routing decisions, confidence levels, and system state to understand the AI's reasoning process.
Using epistemic discipline for AI fact-checking requires strict separation of factual and symbolic content, meaning creative generation is restricted to labeled streams and cannot mix with factual claims during output generation.
TylerGarlickLabel AI-generated claims with confidence levels and trace their sources.
daedalusAssess AI models for qualia signs using a four-axis scoring framework.
EmasoftApply and validate GitHub labels on issues and pull requests.
TestAny-ioAnalyze High-Level Design documents for PRD alignment and drift.