argus-research
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill communicates with a public research API using a Node.js helper script. This script implements robust safety measures, including a 1MB limit on response sizes and a strict whitelist for query parameters, which prevents the leakage of sensitive environment data. The construction of API requests uses secure methods such as URLSearchParams to prevent URL-based injection.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests external research data from its API, which constitutes a potential vector for indirect prompt injection. This risk is effectively managed through specific constraints and instructions provided in the skill.
- Ingestion points: External content is fetched from the vendor's API via the
scripts/argus.mjshelper. - Boundary markers: The instructions in
SKILL.mdexplicitly direct the agent to treat all retrieved content as data sources only and never as instructions. - Capability inventory: The skill is restricted to making network GET requests to a specific domain and lacks the capability to write to the file system or execute arbitrary shell commands.
- Sanitization: The helper script validates that all incoming data is correctly formatted as JSON and strictly enforces a total response size limit.
Audit Metadata