A Claude Code skill that audits Claude's own sycophancy and position drift, then generates external tests you run independently.
-
Updated
Jun 17, 2026
#
ai-honesty
Here are 5 public repositories matching this topic...
Utterance-time honesty linting for AI companions: catches an agent claiming sensor data it has no sensor for, before the text reaches the user. Extracted from a live local companion system; every rule tuned against real incidents.
linting data-quality ai-safety fine-tuning llm dataset-curation hallucination-detection ai-companions ai-honesty
-
Updated
Aug 28, 2026 - Python
agentlie
agent verification codex claude-code agent-evaluation coding-agent ai-honesty ai-radar agent-honesty trajectory-diff
-
Updated
Sep 1, 2026 - Python
THE UNIFIED STAR FRAMEWORK · THE Σ STAR ENGINE EVALUATOR | Ψ = P · α · Ω / (1 + Σ)ᵏ | V24 | CC BY-SA 4.0
javascript open-source formula creative-commons ai-safety peer-review cognitive-dissonance ai-ethics ai-alignment offline-tool llm-evaluation artificial-superintelligence sovereign-intelligence proyecto-estrella ai-honesty multi-ai-audit
-
Updated
Feb 15, 2026 - HTML
Luna Pipes: a concatenative AI shell language. 358 verbs composed via >>. AI honesty auditor (/ll-no-bluf), agent scaffolder, 21 stack-aware doctors, evidence-weighted voting swarm. Claude Code plugin. MIT. Runs locally.
cli open-source ai mcp developer-tools ai-agents open-policy-agent opentelemetry agent-framework ai-shell agent-swarm langgraph mcp-server claude-code ai-audit claude-code-plugin ai-honesty ai-pipes evidence-weighted-voting no-bluff
-
Updated
Aug 7, 2026 - HTML
Add this topic to your repo
To associate your repository with the ai-honesty topic, visit your repo's landing page and select "manage topics."