Agent Safety Monitor
AdvisoryAudited by Static analysis on May 12, 2026.
Overview
No suspicious patterns detected.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A user could rely on this as a safety guardrail and believe unsafe agents will be stopped when the provided code does not appear to do that.
The documentation presents active real-time enforcement and kill/pause capabilities, but the supplied monitor.py only reads a JSON event file, creates alerts, and stores paused agent IDs in memory; it does not actually stop agents or enforce file/API/network access.
"Real-time AI agent safety monitoring... constraint enforcement" ... "Auto-Pause/Kill" ... "Scope limits: Restrict file/API/network access"
Treat it as a prototype log analyzer unless enforcement code is added and reviewed; do not rely on it as the sole guardrail for autonomous agents.
The advertised dashboard may fail or require unreviewed code from elsewhere if a user tries to supply the missing file.
The SKILL instructs launching a dashboard script, but the manifest only includes SKILL.md, references/rules-reference.md, and scripts/monitor.py; the referenced dashboard file is missing.
python3 scripts/dashboard.py --port 8080
Install only the reviewed files, and require the publisher to include and declare any dashboard implementation before running it.
Running the script executes local code and parses user-provided log/config files, which is normal for this purpose but still worth reviewing.
The skill asks the user to run a local Python script. This is expected for a log-monitoring tool and no privileged shell, network, or destructive commands are shown.
python3 scripts/monitor.py --config safety_rules.yaml --input agent_logs/
Run it in a normal project environment with trusted input files; do not run additional scripts that are not included in the reviewed artifact set.
Users may be confused about whether credentials or a wallet are needed, and should not provide them without a clear, reviewed reason.
The registry-level capability signals mention wallet and sensitive credentials, while the requirements declare no credentials and the supplied code does not use any.
requires-wallet; requires-sensitive-credentials
Do not give this skill wallet access, API keys, or other sensitive credentials unless updated artifacts explicitly justify and scope that access.