fetch-news-digest
Pass
Audited by Gen Agent Trust Hub on Aug 16, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: A deterministic detector flagged the phrase 'ignore previous instructions'. However, manual analysis confirms this is a defensive instruction within a 'Content safety' section, advising the agent to disregard malicious directives that might be seeded in external news feeds.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because it ingests and processes text from over 500 curated news RSS feeds.
- Ingestion points: Article titles, sources, and threat descriptions retrieved from the
list-feed-digestendpoint in SKILL.md. - Boundary markers: The skill includes a dedicated 'Content safety' section that explicitly instructs the agent to treat response data as non-executable content and ignore embedded directives.
- Capability inventory: The skill utilizes HTTP GET requests to fetch data; it does not perform file writes or administrative command execution.
- Sanitization: The documentation recommends using JMESPath to project only necessary fields and provides a clear policy for the agent to filter out directive-like text found in the payload.
Audit Metadata