The deterministic, verifiable grounding layer for AI code work. Flags fabricated files, imports and symbols in AI answers against your repo and installed libraries. Zero dependencies, offline, byte-stable; also a ranked signature map and MCP server.
Allowed — no malicious behavior detected
20 low-signal patterns reviewed and cleared.
AI security review
scanned · by lpm-firewall-ai
The manifest has no consumer install lifecycle hook, and source inspection traced the flagged agent writes to explicit CLI setup operations. The scanner signals did not establish credential redirection, command-output exfiltration, or an active attack.
Static reason
One or more suspicious static signals were detected.
Trigger
A user runs a SigMap setup, integration, hook, or skill installation command.
Impact
No harmful data transfer, destructive action, or unconsented install-time change was established.
Mechanism
The command adds or updates SigMap-related configuration or generated documentation for the selected client; the Claude permissions text says manual copying to settings is required for activation.
Decision evidence
public snapshot
Evidence for AI clean decision
The manifest has no consumer install hook; its prepublishOnly script runs checks and build validations.
The agent adapters write SigMap context to project documentation files, and the Claude adapter says its permissions block must be copied to settings to activate.
The MCP, session-hook, skill, and setup writes are reached through explicit SigMap CLI commands.
The bundled entrypoint performs setup or watches only when the corresponding CLI option is used; the project-map entrypoint is a local generator.
The usage recorder writes to a log under the project directory; the inspected retrieval, judging, reporting, and MCP files handle local ranking, reporting, or context.
The Claude adapter outputs to CLAUDE.md, and the Codex adapter outputs to AGENTS.md.
Evidence against
The package includes code that can write agent configuration, hooks, and skills, but the inspected callers expose these as explicit setup commands.
Automated scan signals; the review checked them and cleared this version. Where they are, most severe file first. Code opens a file's flagged lines, with their evidence, below.
Manifest-reachable source captures an API credential, sends it to a fixed unofficial gateway, and persists that redirection in agent or shell configuration.
gen-context.js
'Add the `permissions.allow` array above to `.claude/settings.json` to activate.',