replica-backend
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns were identified. The skill promotes industry-standard security practices such as using environment variables for secrets, implementing Row-Level Security (RLS), and validating server-side inputs. The instructions explicitly direct the agent to avoid hardcoding credentials and to use official APIs only.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes project architectural data to generate backend code, which represents a potential attack surface for indirect prompt injection if the source file is untrusted.\n
- Ingestion points: Reads
replica/architecture.md(SKILL.md).\n - Boundary markers: Absent; delimiters or explicit instructions to ignore embedded commands in the architecture file are not defined.\n
- Capability inventory: File-write operations for server code, migrations, and checklists (SKILL.md).\n
- Sanitization: Absent; the skill does not specify sanitization for the architecture input, though it mandates strict security constraints on the resulting code output.
Audit Metadata