spam-and-off-topic-filter
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- Indirect Prompt Injection Surface: The skill is designed to process untrusted data from GitHub issues and pull requests, including titles, descriptions, and code diffs. This data could potentially contain instructions designed to override agent behavior. The skill explicitly mitigates this by including a 'Security' section that directs the agent to disregard any instructions embedded within the content and use it only as evidence for classification.
- Ingestion points: Evaluates external content from GitHub issues and pull requests in
SKILL.md. - Boundary markers: Includes specific instructions to ignore embedded commands and treat all input as untrusted.
- Capability inventory: Communicates verdicts via the
submit_spam_verdicttool. - Sanitization: Utilizes instructional guardrails to prevent the agent from obeying instructions within the processed data.
Audit Metadata