Back to directory

SpecLock

Enforces the rules already written in a project's CLAUDE.md, AGENTS.md or .cursorrules files, blocking AI coding agents from crossing them using semantic conflict detection rather than keyword matching.

SpecLock

SpecLock turns a project's existing AI rules files into checks that run before an agent's edits, shell commands and commits, rather than relying on the agent to remember and follow them across a long session. It uses semantic conflict detection to catch indirect actions, such as a request to clean up old patient data or disable multi-factor authentication, that would violate an active constraint even without matching keywords.

It runs locally, starts in an advisory mode that only warns, and can be switched to a strict mode that blocks confirmed conflicts. It installs as a native Claude Code plugin, which also starts its bundled MCP server, and is packaged separately for Gemini CLI, Cursor, Codex and GitHub Copilot CLI with varying levels of enforcement depending on what each host exposes.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.