codebase-architecture
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides architectural guidance and security best practices for developing and maintaining TypeScript codebases.\n- [INDIRECT_PROMPT_INJECTION]: The skill includes explicit security measures to prevent indirect prompt injection.\n
- Evidence: In
references/agent-runtime.md, the skill recommends using delimited data blocks for untrusted inputs (logs, issue reports) and stripping potential escape sequences to ensure the agent does not execute instructions embedded in data.\n- [CREDENTIALS_SAFE]: Security guidance advocates for scoped user credentials and branded IDs, preventing privilege escalation and accidental data exposure.\n - Evidence: Found in
references/api-design.mdandreferences/distributed-correctness.md.\n- [COMMAND_EXECUTION]: Instructions for using standard development tools (e.g.,lint,test,knip,jscpd) are provided in the context of enforceable repository guardrails.\n - Evidence: The skill presents these within a security framework that includes per-run spend caps, isolated settings, and non-interactive permission modes in
references/agent-runtime.md.\n- [PRIVILEGE_ESCALATION]: The skill mandates a 'least privilege' approach for agents.\n - Evidence: Recommends separating read-only analysis jobs from write-capable jobs and requiring human review for all merges in
references/agent-runtime.mdandreferences/enforcement-ladder.md.
Audit Metadata