slack-pdf-summary
Pass
Audited by Gen Agent Trust Hub on Sep 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Dart script located at
/Users/mono/Git/script/bin/script.dartusingdart runor ataskcommand. This execution incorporates user-provided inputs such as the PDF file path and optional arguments like channel names and titles. - [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted content from PDF files which are then provided to the Gemini AI model for summarization, creating a surface for indirect prompt injection attacks.
- Ingestion points: Content of PDF files specified by the user or provided as attachments.
- Boundary markers: None identified in the skill instructions to separate the data from the processing instructions or to warn the model against following instructions embedded within the document.
- Capability inventory: The skill has capabilities to read local files, interact with the Gemini API for content processing, and use the Slack API for file uploads and message posting.
- Sanitization: There is no mention of sanitization, filtering, or validation of the PDF content before it is passed to the AI model or the Slack API.
Audit Metadata