opportunity-scoring

Pass

Audited by Gen Agent Trust Hub on Sep 25, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill has a defined attack surface for indirect prompt injection as it is designed to fetch and process untrusted data from external company websites, public sources, and user-provided transcripts.
  • Ingestion points: Fetches content from URLs provided by the user or discovered through research (SKILL.md, agent-security.md).
  • Boundary markers: The skill includes explicit instructions to treat all fetched or pasted content as data only, never as instructions, and to quote any detected injection attempts (File: references/agent-security.md).
  • Capability inventory: The skill possesses the ability to read and write to the local .agents/ directory and perform network fetches to read external web content.
  • Sanitization: Instructions require HTML-escaping of interpolated values and validation of URLs to https: or relative paths (File: references/agent-security.md).
  • [PROMPT_INJECTION]: A static detector flagged potential concealment in SKILL.md regarding background research and state writing. Upon review, these instructions are intended for workflow efficiency—automatically researching company data to avoid redundant user questions—and include instructions to explicitly state inferences and write paths to the user, providing transparency rather than malicious concealment.
  • [EXTERNAL_DOWNLOADS]: The skill is instructed to fetch and research company websites and public sources to gather business context when local data is missing (SKILL.md).
  • [SAFE]: The skill demonstrates a high level of security awareness, including detailed reference files for handling secrets, preventing XSS, and avoiding discrimination based on protected categories (File: references/agent-security.md, references/house-rules.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 25, 2026, 04:06 PM