diffx-finish-review

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill fetches comment bodies from a local API and instructs the agent to apply the changes described in that text. If the comments contain malicious instructions, the agent might execute them as if they were valid user requests.\n
  • Ingestion points: SKILL.md (via curl http://localhost:<port>/api/comments)\n
  • Boundary markers: The instructions lack explicit delimiters or warnings to ignore instructions embedded within the comment body.\n
  • Capability inventory: curl for network requests and file system modification (implied by the instruction to apply changes to files at filePath).\n
  • Sanitization: No sanitization or validation of the external comment content is performed before interpretation by the agent.\n- [COMMAND_EXECUTION]: The skill provides templates for shell commands using curl to interact with a local API. While targeting localhost, these commands represent a capability that could be misused if the agent's context is manipulated by malicious data.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 02:57 AM
Security Audit — agent-trust-hub — diffx-finish-review