skill-creator
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script,
scripts/validate-metadata.py, via the command line to validate user-supplied metadata.\n- [INDIRECT_PROMPT_INJECTION]: The skill's instructions require the agent to interpolate user-provided strings (nameanddescription) directly into a shell command. This creates a vulnerability surface where crafted user input containing shell metacharacters could lead to arbitrary command execution.\n - Ingestion points: The
[name]and[description]placeholders inSKILL.mdare intended to be replaced with user-provided content.\n - Boundary markers: No delimiters or escaping instructions are provided in the prompt to protect against shell injection.\n
- Capability inventory: The agent is expected to execute local scripts using
python3.\n - Sanitization: The
scripts/validate-metadata.pyfile performs its own validation (regex and character checks), but the shell command execution itself is initiated before the script's internal logic runs, leaving the invocation step vulnerable.
Audit Metadata