java-agent-instructions
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill manages the agent's persistent instruction context by modifying configuration files. \n
- Ingestion points: Reads the content of
AGENTS.mdorCLAUDE.mdfrom the project root inscripts/install-agent-instructions.py. \n - Boundary markers: Uses explicit HTML-style comments
<!-- java-skills:self-review:start -->and<!-- java-skills:self-review:end -->to delimit the injected block. \n - Capability inventory: The script performs file read and write operations on project-level instruction files. \n
- Sanitization: The script performs structural validation of markers (e.g., checking for malformed or reversed tags) before committing changes to ensure file integrity. \n- [COMMAND_EXECUTION]: The test suite
tests/test_installer.pyutilizessubprocess.runto execute the local installer script. This is a standard testing practice for CLI tools and is restricted to the test environment using static paths to the skill's own assets. \n- [SAFE]: The skill's primary function is to help maintain project-level documentation. All logic is local, and it uses a static template from theassets/directory rather than fetching instructions from external remote sources.
Audit Metadata