summarize

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data from URLs or local files and interpolates the resulting Markdown content directly into a prompt for the pi summarization tool.
  • Ingestion points: The input argument in to-markdown.mjs accepts arbitrary URLs and file paths, loading their content into the agent's context.
  • Boundary markers: The summarization prompt uses --- BEGIN DOCUMENT (Markdown) --- and --- END DOCUMENT --- delimiters to separate the document content from instructions.
  • Capability inventory: The skill has the ability to execute subprocesses (uvx, pi) and write files to the local filesystem (e.g., /tmp/pi-summarize-out/).
  • Sanitization: There is no evidence of sanitization or escaping of the ingested document content before it is interpolated into the summarization prompt, which could allow an attacker-controlled document to override instructions.
  • [COMMAND_EXECUTION]: The to-markdown.mjs script uses child_process.spawnSync to invoke external CLI tools including uvx and pi. While the script safely uses argument arrays to prevent shell injection, it facilitates the execution of external binaries based on user-provided inputs.
  • [EXTERNAL_DOWNLOADS]: The skill leverages Microsoft's markitdown utility to fetch content from remote URLs and uses uvx to dynamically install the package from PyPI during execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 02:31 PM