explain-codebase
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is designed for architectural analysis and explanation with no identified malicious patterns. It explicitly mandates a read-only boundary, instructing the agent not to modify code or documentation. No network operations, credential access, or privilege escalation commands are present in the skill files.
- [NO_CODE]: The skill consists entirely of Markdown instructions and YAML metadata. It does not include any executable scripts, binary files, or external dependencies that could facilitate remote code execution.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted codebase files, which constitutes an attack surface for indirect prompt injection. 1. Ingestion points: The agent is instructed to read source code, repository searches,
CONTEXT.md, and architectural decision records (ADRs) to trace system logic. 2. Boundary markers: The skill includes an explicit 'read-only' directive and instructions to distinguish observed behavior from inferred rationale. 3. Capability inventory: The skill has no capabilities for file writing, network communication, or subprocess execution. 4. Sanitization: While no technical sanitization of file content is specified, the instructional constraints focus the agent on objective observation rather than executing instructions found within the data.
Audit Metadata