skill-management

Warn

Audited by Gen Agent Trust Hub on May 19, 2026

Risk Level: MEDIUMREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill uses npx to download and execute packages from arbitrary, user-specified repositories (npx --yes skills add <owner/repo>), which can lead to the execution of untrusted code.\n- [EXTERNAL_DOWNLOADS]: The skill performs git clone operations on repository URLs provided by the user, downloading external code into the agent's environment without integrity verification.\n- [PROMPT_INJECTION]: The skill instructions for 'Same-Session Use' create a significant surface for indirect prompt injection (Category 8).\n
  • Ingestion points: Untrusted SKILL.md files imported from external GitHub repositories into the .claude/skills/ directory.\n
  • Boundary markers: Absent; the skill explicitly recommends injecting the raw content at the top of the task prompt.\n
  • Capability inventory: Includes shell command execution via npx, git, rm, cp, and python3, as well as file system modification.\n
  • Sanitization: Absent; no validation or escaping of the imported markdown content is performed before injection.\n- [COMMAND_EXECUTION]: The skill executes multiple shell commands (find, cp, rm, python3) that process paths and names derived from external sources, posing a risk of command injection if the recommended validation steps are bypassed.
Audit Metadata
Risk Level
MEDIUM
Analyzed
May 19, 2026, 10:57 AM
Security Audit — agent-trust-hub — skill-management