show-me
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the current conversation to generate visual artifacts, creating a potential injection surface.
- Ingestion points: The instructions state to "Help the user understand the current topic of conversation visually," which involves processing arbitrary content from the chat history (SKILL.md).
- Boundary markers: There are no explicit instructions to use delimiters or sanitization logic when interpolating conversation data into the generated visual artifacts.
- Capability inventory: The skill utilizes file-writing capabilities and shell execution (Bash) to create and display HTML files.
- Sanitization: The skill lacks instructions to sanitize or escape external content before it is placed into HTML files or used in shell command templates.
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command to open generated visual artifacts.
- Evidence: The instruction to use
Bash(open path/to/show-me-{description}.html)involves runtime shell execution. While intended for opening local files, a lack of strict validation on the{description}variable could theoretically be targeted for command injection if the agent is not sufficiently constrained by the platform.
Audit Metadata