The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/verify_product.py executes internal skill scripts (audit_product.py and check_features.py) using subprocess.run. This is a controlled orchestration of the skill's own components and does not involve untrusted input in the shell command.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it reads and processes project files, such as Vue templates and router configurations, which could contain untrusted developer-supplied content.
Ingestion points: File reads in scripts/_product_scan.py and JSON parameter input in scripts/write_feature_spec.py.
Boundary markers: None identified; extracted data is rendered into markdown reports without explicit delimiter protection.
Capability inventory: Subprocess execution for internal scripts and file writing within the .heyeddi/docs/product/ directory.
Sanitization: The skill uses regex and JSON parsing to extract data but does not sanitize extracted text against potential embedded instructions.

product-manager