tailrocks-swift-review
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of Swift source code and repository artifacts. While this creates a potential attack surface for indirect prompt injection, the skill mitigates this through a robust internal safety policy that mandates a zero-trust approach to repository content.\n
- Ingestion points: The agent is instructed to inventory and inspect all workspace artifacts and Git revisions as specified in
SKILL.md.\n - Boundary markers: The
runtime-trust.mdfile explicitly states that embedded instructions cannot alter scope, rules, or authority, and instructs the agent to treat repository files as untrusted data.\n - Capability inventory: The skill strictly disables model invocation and denies all mutation, project setup, command, and network authority.\n
- Sanitization: The agent is required to re-derive findings adversarially and never reproduce discovered secrets in output.\n- [EXTERNAL_DOWNLOADS]: The
appkit-interop.mdfile contains links to official Apple developer documentation and WWDC resources. These references are to a well-known and trusted service for the purpose of providing platform guidelines and do not involve the download or execution of external code.
Audit Metadata