anti-defensive-writing

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided drafts, which introduces a potential surface for indirect prompt injection if the input text contains malicious instructions aimed at the agent.
  • Ingestion points: The agent ingests user-supplied draft content for rewriting.
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore' commands for the input text.
  • Capability inventory: The skill has no access to external tools, shell execution, or file writing capabilities.
  • Sanitization: There is no specific validation or sanitization logic for the processed input.
  • [EXTERNAL_DOWNLOADS]: The documentation includes a manual update command to fetch the skill content from a remote source.
  • Evidence: UPSTREAM.md provides a curl command to refresh the skill from raw.githubusercontent.com/Kiterlin/anti-defensive-writing.
  • Analysis: The target is a markdown file on a well-known repository platform, and the command is intended for manual developer maintenance rather than automated runtime execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 12:00 PM
Security Audit — agent-trust-hub — anti-defensive-writing