software-engineering-guidelines
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill requires the agent to execute the project's test suite, linter, and type checker, which involves executing arbitrary commands defined in the project's own configuration files.\n
- Evidence: Principle 4 states that the agent must "Run the full test suite according to the project's rules" and "Run the project's linter and type checker when provided".\n- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from user requests and project source code, using it to inform file-writing and command-execution tasks without explicit sanitization or boundary markers.\n
- Ingestion points: User requests, project source code, and agent-facing documentation files like AGENTS.md.\n
- Boundary markers: There are no specific delimiters or instruction-ignore warnings defined to isolate instructions from external data.\n
- Capability inventory: The agent is authorized to write to source files and documentation, and to execute shell commands for testing and linting.\n
- Sanitization: The instructions do not specify a process for sanitizing, escaping, or validating external content before it is processed or written to files.
Audit Metadata