systematic-debugging
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an inherent attack surface for indirect prompt injection due to its core function of processing untrusted codebase files and external documentation.
- Ingestion points: The workflow involves inspecting local repository files, runtime inputs, and external documentation retrieved via
sandbase_inspect(SKILL.md, references/sandbase-api-map.md). - Boundary markers: The instructions explicitly direct the agent to "Preserve the user's chosen... authorization boundaries" and "Send only the minimum necessary data" (SKILL.md).
- Capability inventory: The skill possesses the ability to read local files and execute external capabilities via the
sandbase_runtool (SKILL.md, references/sandbase-api-map.md). - Sanitization: The skill mitigates risks through mandatory quality gates, including "Verify the real output with appropriate tests," "Base conclusions on a reproducible failing or passing check," and requiring user confirmation for all paid or external tool executions (SKILL.md).
Audit Metadata