subgraph-doctor
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs diagnostic and administrative actions using the goldsky CLI tool.
- Evidence: Commands include goldsky project list, goldsky subgraph list, goldsky subgraph log, goldsky subgraph start, goldsky subgraph pause, goldsky subgraph deploy, and goldsky subgraph delete found in SKILL.md.
- [INDIRECT_PROMPT_INJECTION]: The skill is exposed to untrusted data through subgraph logs which are processed to determine fixes.
- Ingestion points: Logs are read from the goldsky subgraph log command in SKILL.md.
- Boundary markers: No delimiters are used to isolate the log content from the agent's instructions.
- Capability inventory: The skill can delete subgraphs or deploy new code, creating a path for unauthorized actions if the agent is misled by log content.
- Sanitization: Log output is analyzed directly without filtering or sanitization steps.
Audit Metadata