quarterly-review
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests Pull Request titles and descriptions from GitHub to synthesize a narrative report. This creates a surface for indirect prompt injection where malicious content in external PR titles could attempt to influence the agent's summary output. \n
- Ingestion points: GitHub PR titles and repository data fetched via the gh CLI (SKILL.md).\n
- Boundary markers: None explicitly mentioned in the synthesis instructions.\n
- Capability inventory: Bash command execution for data fetching (gh) and processing (jq), combined with file system Read and Write for temporary storage.\n
- Sanitization: Content is filtered through jq and synthesized by the LLM, but no specific sanitization for malicious instructions within titles is documented.- [SAFE]: The skill interacts exclusively with GitHub, a well-known and trusted service, using official tooling (gh). All data processing is performed using standard utilities (jq) on local temporary files, with no evidence of exfiltration to untrusted domains or unauthorized privilege escalation.
Audit Metadata