penetration-testing

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (ls) to check for the presence of recommended companion skills within the local file system (e.g., ~/.claude/skills/).
  • [EXTERNAL_DOWNLOADS]: The instructions suggest downloading and installing remote components via the npx skills add command from the AbsolutelySkilled repository.
  • [PROMPT_INJECTION]: The skill is a surface for indirect prompt injection (Category 8c) as it processes untrusted data from external targets during security assessments.
  • Ingestion points: According to SKILL.md and references/methodology.md, the agent processes output from network scanners (Nmap), web fuzzers (ffuf), and remote server responses (HTTP headers and HTML body content).
  • Boundary markers: There are no instructions to use delimiters or ignore embedded commands within the ingested tool outputs.
  • Capability inventory: The agent has the capability to execute shell commands (Nmap, ffuf, ls) and write to the file system (generating reports) as described across the skill files.
  • Sanitization: The skill does not provide mechanisms for validating or sanitizing the data received from external tools before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 03:05 PM
Security Audit — agent-trust-hub — penetration-testing