The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the workspace and implementation plans, which could contain malicious instructions designed to influence the agent's behavior during the audit process.\n
Ingestion points: The audit_engineering.py script reads source files from src/ and backend/ directories, and check_engineering_plan.py accepts external plan text or files for analysis.\n
Boundary markers: The skill does not employ specific delimiters or warnings to indicate that the processed code or plans should be treated as untrusted data.\n
Capability inventory: The skill is capable of writing files to the .heyeddi/ directory via append_decision.py, audit_engineering.py, and init_engineering_docs.py. Although a subprocess execution helper is present in _skill_cli.py, it is not utilized by the core logic of the audit tools.\n
Sanitization: There is no evidence of sanitization or filtering of the content read from external files before it is processed by the agent.\n- [COMMAND_EXECUTION]: The skill includes boilerplate for command execution and instructions for the agent to run scripts from related skills.\n
Evidence: The _skill_cli.py file contains a run_command helper that uses subprocess.run with shell=False to execute commands.\n
Evidence: SKILL.md contains instructions for the agent to execute a script (suggest_next_skill.py) located in a sibling skill directory (heyeddi-orchestrator) for workflow handoffs.

engineering-excellence