app-copy

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its core functionality of reading and rewriting UI strings.
  • Ingestion points: The skill scans JSX/TSX files, toast/notification messages, error boundaries, and empty state components (SKILL.md Phase 1).
  • Boundary markers: None identified. The agent is instructed to read strings directly without delimiters or instructions to ignore embedded commands.
  • Capability inventory: The skill can write rewritten strings directly back into the codebase (SKILL.md Phase 7).
  • Sanitization: No explicit sanitization, validation, or escaping of ingested strings is performed before they are used to generate new output.
  • Autonomy Abuse: The instructions state, "If the codebase tells you, don't ask," which encourages autonomous action based on potentially untrusted data found within the source code.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 10:12 AM
Security Audit — agent-trust-hub — app-copy