Caveat verdict

Agent Browser

lobster-training

88
🟢 Trusted
No high-risk patterns surfaced by the deep scan — automated capability review, not behavioral proof.

Headless browser automation CLI (agent-browser) for navigating and interacting with pages; capabilities listed are standard browser automation operations with no evidence of secret exfiltration or remote code execution.

⚠ Flagged for review — coarse, uncorroborated signal, not a confirmed exploit. Review the config yourself before installing.

Automated static analysis — not a human review. Caveat flags capabilities, not confirmed intent, and can produce false positives. Disagree with this verdict? Use Dispute below.

48
security
70
transparency
70
maintenance

Permission integrity

Installs packages at runtime — transitive dependencies are not auditable

package_install

Findings (5)

Pattern match medium

References agent configuration files

references/examples.md · code · CLAUDE.md

Pattern match medium

References agent memory files

lobster-training-skill/SKILL.md · code · MEMORY.md

Pattern match medium

References child_process — can spawn system processes

lobster-training-skill/scripts/training.js · prose · downgraded · child_process

Pattern match medium

References sudo — requests elevated privileges

skill-card.md · prose · downgraded · sudo

Pattern match low

Python os.environ.get — reads environment variable

scripts/generate_image.py · prose · downgraded · os.environ.get(

Why the tier is capped

Execution sink present in raw bytes (Hard Floor: class B/D). Final tier capped at Caution — cannot be lifted by any downgrade, example-payload opt-in, or allowlist.

Permissions & capabilities

Requires 2 system binaries. (1 elevated: npm).

package_installdynamic_eval

Is this flag fair?

Check another skill Browse the registry Auditing your own skills or configs? Use the API