The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided BRAINSTORM_.md files to extract requirements and generate documentation. 1. Ingestion points: BRAINSTORM_.md files are read as the primary input for the Define phase. 2. Boundary markers: Absent; the skill does not explicitly instruct the agent to ignore potential instructions embedded within the input markdown files. 3. Capability inventory: The skill can write files to the local filesystem and execute a local Python validation script. 4. Sanitization: Content is mapped into a structured format using assets/DEFINE_TEMPLATE.md, which helps isolate data from instructional logic.
[COMMAND_EXECUTION]: The skill executes a bundled Python script (scripts/validate_define.py) to perform structural linting on the generated requirement documents. The script is static, provided within the skill package, and performs regex-based text parsing. There is a minor potential for argument injection if the agent generates a malicious filename from user input, though the script itself is safe.

sdd-define