audit-backend-architecture

Warn

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill is instructed to locate sensitive connection information by searching through configuration files, including .env files.
  • Evidence: In references/patterns.md, Pattern 19 uses the command rg -n "schema|search_path|DATABASE_URL|datasource" -g '*.{ts,js,py,prisma,toml,env}' -l to identify where database URLs are defined, targeting files known to store secrets.
  • [INDIRECT_PROMPT_INJECTION]: The skill audits and interprets untrusted source code and configuration files, creating a vulnerability where malicious instructions embedded in the code could influence the agent's behavior.
  • Ingestion points: The skill uses ripgrep to scan various files including .ts, .js, .py, .go, .sql, .yaml, and .env across the codebase (defined in Phase 0 and references/patterns.md).
  • Boundary markers: While the skill specifies it is "Read-only" and requires "evidence," there are no explicit delimiters or instructions provided to the agent to treat code content as untrusted data or to ignore embedded natural language instructions.
  • Capability inventory: The agent has the capability to read any file matching the audit patterns and access specific lines using rg and fd commands.
  • Sanitization: No sanitization or filtering is performed on the content retrieved from the audited files before it is processed by the AI agent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 7, 2026, 11:26 PM