skill-best-practices
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled Python script
scripts/validate-metadata.pyto validate metadata. While the script itself is a benign validation tool, executing scripts with user-controlled parameters is a privileged operation. - [DYNAMIC_EXECUTION]: The skill instructions (Step 4 and Step 5) and the provided template (
assets/SKILL.template.md) encourage the generation and execution of new scripts within thescripts/directory to handle specific tasks. This is a standard part of the skill's utility but represents a surface for dynamic code execution. - [INDIRECT_PROMPT_INJECTION]: The procedural step in
SKILL.mdinterpolates user-supplied data ([name]and[description]) directly into a shell command. - Ingestion points: Metadata values (name and description) provided by the user during the skill creation process.
- Boundary markers: None present; the values are wrapped in double quotes in the instruction but rely on the agent's internal logic for proper escaping.
- Capability inventory: Execution of Python scripts via the shell.
- Sanitization: No validation or escaping logic is defined in the instructions to prevent shell metacharacters from triggering command injection.
Audit Metadata