{"schema_version":1,"snapshot_id":"bfb936f5ceae4a7427f93961201723779c3021545a3c4222395e8678314c3e55","generated_at":"2026-10-17T21:51:24.907033+00:00","notice":"Potential GTM matches from directory metadata and model triage. Discovered does not mean vendor-verified or tested. Pending records remain under review; pricing and install details require primary sources.","canonical_url":"https://gtmmcps.com/discoveries/52e2e224bad21be80abd","record":{"id":"52e2e224bad21be80abd","name":"agent-guard","aliases":["agent-guard"],"description":"  **The safety check an AI agent runs *before* it acts.**\n\n Every agent that installs packages, runs shell commands, or commits/outputs text can do real damage — install\nmalware, wipe a disk, or leak a credential. agent-guard is the cheap \"look before you leap\" gate an agent calls before the irreversible step.\n\n  | tool | call it before… |\n  |---|---|\n  | `check_package` | `pip/npm install` — is it real, malware, a typosquat, or does it run code on install? |\n  | `check_command` | running a shell command — destructive\n  / RCE (rm -rf, curl\\|bash, dd, force-push)? |\n  | `scan_secrets` | committ","repository_url":null,"sources":[{"source_id":"smithery","url":"https://smithery.ai/servers/ipezygj2/agent-guard","observed_at":"2026-10-17T21:40:22.481400+00:00","source_record_id":"89150e73-1ddd-451b-b168-cee28b00b43d","source_record_name":"ipezygj2/agent-guard"}],"relevance_signals":["payment"],"classification":{"label":"pending","method":"not-yet-classified"},"catalog_ids":[],"verification_status":"discovered","vendor_ownership":"not-verified","execution_tested":false,"pricing_status":"not-verified","documentation_status":"not-verified"}}