doc-compression

Pass

Audited by Gen Agent Trust Hub on May 11, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructions provide a path for the agent to install additional PDF extraction capabilities from a trusted repository (github.com/anthropics/skills) if such tools are not already available in the environment. This reference targets a well-known, trusted organization and is documented for legitimate functional extension.
  • [COMMAND_EXECUTION]: The skill includes a local Python script, scripts/word_count.py, which is used to calculate document metrics. It also provides instructions for generating PDF outputs using standard document tools such as Pandoc or HTML-to-PDF renderers, which is typical for a document-processing skill.
  • [PROMPT_INJECTION]: As the skill is designed to process untrusted text documents, it is exposed to potential indirect prompt injection. However, the instructions explicitly mandate a 'compression by elimination' workflow that prohibits the agent from paraphrasing or interpreting the text, effectively acting as a defense-in-depth measure against the execution of instructions embedded within processed documents.
Audit Metadata
Risk Level
SAFE
Analyzed
May 11, 2026, 04:39 PM
Security Audit — agent-trust-hub — doc-compression