competitive-intel-watch
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from untrusted external sources, creating a surface for indirect prompt injection attacks where an attacker could influence the agent's behavior via content on a monitored website.
- Ingestion points: In
SKILL.md, the search plan includes company sites, pricing pages, release notes, news articles, review sites, and job postings. - Boundary markers: The instructions mandate the use of
Fact/Inference/Assumptionlabels and require explicitURLanddateevidence for every claim, which provides some structural grounding against instruction following from ingested data. - Capability inventory: The skill utilizes search and browsing capabilities to aggregate information and provides options to trigger downstream updates to battle cards and roadmap assumptions.
- Sanitization: There are no explicit instructions for technical sanitization of the scraped content; the skill relies on the agent's internal safety guardrails and a logical 'materiality bar' to filter out irrelevant or malicious noise.
Audit Metadata