share-reading

Pass

Audited by Gen Agent Trust Hub on Mar 7, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits vulnerability to indirect prompt injection (Category 8) by processing untrusted data from URLs and local files. \n
  • Ingestion points: The WebFetch tool is used to ingest data from external URLs, and the Read tool is used to ingest data from local file paths. \n
  • Boundary markers: None. The skill does not use delimiters or instructions to ignore embedded commands in the processed content. \n
  • Capability inventory: WebFetch (network read), Read (file system read), and text generation/social media drafting. \n
  • Sanitization: None. The skill does not sanitize the input before interpolation or processing.\n- [COMMAND_EXECUTION]: The skill explicitly instructs the agent to use a Read tool on user-provided file paths. This functionality can be abused to access sensitive local files (e.g., configuration files, credentials, or SSH keys) if the underlying agent environment does not enforce strict path-based access controls or sandboxing.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 7, 2026, 02:32 PM