generate-custom-instructions-from-codebase
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection by analyzing arbitrary codebase content to derive instructions for future AI behavior.
- Ingestion points: The agent is instructed to read and compare source code between different project states (branches, commits, or tags) across the defined project scope in SKILL.md.
- Boundary markers: The prompt fails to provide explicit delimiters or instructions for the agent to ignore or isolate natural language instructions that might be embedded within code comments or documentation in the analyzed repository.
- Capability inventory: The skill utilizes extensive file-read capabilities to perform comparative analysis and file-write capabilities to create a new
.github/copilot-migration-instructions.mdfile. - Sanitization: The instructions do not include any validation or filtering mechanisms to ensure that the extracted 'transformation patterns' or 'emerging patterns' are legitimate code evolutions rather than injected malicious directives.
- [NO_CODE]: The skill does not include any accompanying scripts, binary executables, or external dependencies. It relies exclusively on the markdown-based instructions provided in the SKILL.md file to guide the agent's behavior.
Audit Metadata