identify-iac-repositories
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes metadata from external repositories (such as names, descriptions, and Semgrep finding paths) via the SubImage MCP server tools. This data enters the agent's context and influences its ranking logic, creating a potential surface for indirect prompt injection where malicious metadata could attempt to override agent instructions.
- Ingestion points: Results from
subimageRunCyphercontaining repository names, descriptions, and automated finding paths (SKILL.md, Steps 4, 6, 7, and 9). - Boundary markers: The skill does not mandate the use of specific delimiters (e.g., XML tags or triple-backticks) to isolate external graph data from instructions.
- Capability inventory: The skill can execute read-only Cypher queries and interact with specialized MCP tools provided by the SubImage server (SKILL.md, Prerequisites).
- Sanitization: The skill contains explicit rules to avoid displaying secret or variable values and requires the agent to summarize output rather than reprinting raw data, which limits the potential impact of malicious content (SKILL.md, Rules).
Audit Metadata