The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to generate new AI agent skills (SKILL.md files and directory structures) based on user-provided names and descriptions. This creates a surface for indirect prompt injection, as malicious instructions could be placed in the generated metadata or body. Ingestion points include the user-supplied skill name and description. The capability inventory involves file system operations (creating directories and files) and packaging (creating zip files). No explicit boundary markers or sanitization methods are documented in the skill structure. However, as the skill is currently documentation-focused and lacks implementation code (such as tool calls or scripts), the risk is structural rather than an active vulnerability.
[SAFE]: No findings related to credentials exposure, remote code execution, obfuscation, or persistence were detected. The skill describes legitimate development workflow automation for team collaboration.

skill-share