link-skills
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions direct the agent to execute
findandln -sshell commands to modify the repository's file system structure. - [INDIRECT_PROMPT_INJECTION]: The skill constructs shell commands by interpolating user-provided skill names from the arguments, creating a vulnerability surface.
- Ingestion points: The
[skill names]argument defined in the frontmatter of SKILL.md. - Boundary markers: No delimiters or explicit instructions are provided to the agent to treat the input data as potentially unsafe or to ignore embedded instructions.
- Capability inventory: The skill uses shell command execution (find, ln) and file reading capabilities.
- Sanitization: There is no evidence of input validation, path normalization, or character escaping for the skill names before they are interpolated into shell command strings.
Audit Metadata