LyingDocs

Version updated for https://github.com/KMing-L/lying-docs to version v1.

  • This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

LyingDocs is a GitHub Action designed to ensure alignment between a repository’s documentation, code, and configuration by detecting discrepancies, such as outdated or misleading documentation, undocumented code paths, or mismatches between stated and actual behavior. It automates the auditing process using two agents: Hermes, which analyzes documentation to extract claims, and Argus, which verifies these claims against the actual codebase, producing a structured report of any misalignments. This helps maintain the repository’s trustworthiness for both humans and AI systems, reducing “trust debt” and improving overall reliability.

What’s Changed

What’s New

GitHub Actions CI Integration

  • New composite action (action.yml) — add LyingDocs to any repo with uses: KMing-L/lyingdocs@v1
  • lyingdocs init-ci command — generates a ready-to-commit .github/workflows/lyingdocs.yml with configurable triggers, backend, and authentication
  • Configurable triggers — run on PRs, version tags, cron schedules, or manually via --trigger pr,tag,manual,schedule
  • Manual approval gate — optional --approval flag adds a review step using GitHub Environment protection rules
  • Automatic PR comments — posts findings as a PR comment with collapsible report; updates in place on re-runs (no spam)
  • Auto model selection — picks gpt-5.4 for OpenAI backends and claude-sonnet-4-6 for Anthropic/Claude backends by default

Anthropic Provider Support

  • Hermes and Argus now support Anthropic as an LLM provider — set provider = "anthropic" in config or use HERMES_PROVIDER / ARGUS_PROVIDER env vars
  • Unified LLM abstraction (llm.py) — OpenAI and Anthropic clients share a common LLMResponse / ToolCall interface with automatic message format conversion
  • Single-key Anthropic pipeline — when using backend: claude_code, both Hermes and Argus use ANTHROPIC_API_KEY by default
  • Claude Code OAuth support — Pro/Max subscribers can use --claude-oauth to run Argus on subscription quota instead of per-API-call billing
  • Custom base URLs — both providers support custom endpoints for proxies and private deployments

Documentation

Upgrade

pip install --upgrade lyingdocs