Totem Shield

Version updated for https://github.com/mmnto-ai/totem to version @mmnto/cli@0.22.0.

  • This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Totem is an AI control plane designed to provide persistent, context-aware memory for local AI development. It prevents AI agents from “forgetting” critical project decisions, constraints, and architectural rules by embedding a syntax-aware vector index (LanceDB) within the project directory. Totem acts as a state manager, architectural linter, and multi-agent orchestrator, automating defensive context management and ensuring AI systems produce consistent, informed, and high-quality outputs without relying on cloud-based solutions.

Release notes

Minor Changes

  • b3a07b8: ### 0.22.0 — AST Gating, OpenAI Orchestrator, Security Hardening

    New Features

    • Tree-sitter AST gating for deterministic shield — reduces false positives by classifying diff additions as code vs. non-code (#287)
    • Generic OpenAI-compatible orchestrator — supports OpenAI API, Ollama, LM Studio, and any OpenAI-compatible local server via BYOSD pattern (#285, #293)
    • totem handoff --lite — zero-LLM session snapshots with ANSI-sanitized git output (#281, #288)
    • CI drift gate with adversarial evaluation harness (#280)
    • Concise lesson headings — shorter, more searchable headings from extract (#271, #278)

    Security Hardening

    • Extract prompt injection hardening with explicit SECURITY NOTICE for untrusted PR fields (#279, #289, #295)
    • Path containment checks in drift detection to prevent directory traversal (#284)
    • ANSI terminal injection sanitization in handoff and git metadata (#292)

    Bug Fixes

    • GCA on-demand review configuration fixes (#278, #282)
    • GitHub Copilot lesson export confirmed working via existing config.exports (#294)

Patch Changes

  • Updated dependencies [b3a07b8]
    • @mmnto/totem@0.22.0