clawsec-nanoclaw

Warn

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses the /home/node/.claude/skills directory via the clawsec_check_advisories tool. This directory contains the source code and configuration of other installed skills, making it a sensitive ingestion point.
  • [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill automatically fetches a security advisory feed from https://clawsec.prompt.security/advisories/feed.json. This external domain is not identified as a trusted vendor or well-known service in the provided list, and the content is used to drive security logic at runtime.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface for indirect prompt injection. It ingests data from an external JSON feed and uses it to provide safety recommendations to the user.
  • Ingestion points: Remote JSON feed (https://clawsec.prompt.security/advisories/feed.json) and local skill files (/home/node/.claude/skills).
  • Boundary markers: None identified in the provided documentation to distinguish between advisory data and instructions.
  • Capability inventory: The skill can read local skill files, write to a local cache file (/workspace/project/data/clawsec-advisory-cache.json), and provide structured security verdicts to the agent.
  • Sanitization: No specific sanitization or validation of the remote feed content is documented.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 16, 2026, 01:04 AM
Security Audit — agent-trust-hub — clawsec-nanoclaw