fusion-developer-app
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a development assistant that reads, reviews, and modifies codebase files, which exposes it to an indirect prompt injection attack surface if files within the repository contain malicious instructions.
- Ingestion points: Untrusted source code files are loaded into the agent context via
file_pathsinputs and directory exploration specified acrossSKILL.md,agents/code-quality.md,agents/data-display.md,agents/design.md,agents/framework.md,agents/styling.md, andagents/testing.md. - Boundary markers: Absent. The instructions lack explicit delimiters or safety commands directing the agent to isolate and ignore embedded natural language prompts within user repository files.
- Capability inventory: No executable scripts, network utilities, file-system writes, or subprocess/eval operations exist within the skill package itself. Its functions are restricted to reading workspace files and generating text/code recommendations.
- Sanitization: Absent. The skill contains no logic to escape, sanitize, or filter out natural language instructions found inside code blocks or comments before processing them.
Audit Metadata