skills/mukul975/anthropic-cybersecurity-skills/performing-oauth-scope-minimization-review/Gen Agent Trust Hub
performing-oauth-scope-minimization-review
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from Microsoft Graph API responses, which constitutes a potential surface for indirect prompt injection.
- Ingestion points: Service principal and OAuth grant data are fetched from Microsoft Graph API endpoints in
scripts/agent.pyandSKILL.md. - Boundary markers: The processing logic lacks explicit delimiters or instructions for the agent to ignore instructions that might be embedded in application metadata.
- Capability inventory: The skill includes functionality in
SKILL.mdto modify or delete OAuth grants. - Sanitization: Data returned from the API is processed as structured JSON but is not specifically sanitized for instruction-like strings.
- [DATA_EXFILTRATION]: The skill communicates with Microsoft Graph API endpoints (
graph.microsoft.comandlogin.microsoftonline.com) to retrieve and manage identity metadata. This represents standard communication with well-known, trusted services for administrative purposes.
Audit Metadata