The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script resolve_customization.py using uv run during both activation and the final handoff step.
[DYNAMIC_EXECUTION]: At Step 12, the skill instructs the agent to treat the output of the configuration resolver for the workflow.on_complete key as a terminal instruction to be executed. This pattern allows for arbitrary shell command execution if the project's configuration files are modified.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from a wide variety of external project files including sprint status YAML, Markdown story records, and architectural documentation. It lacks clear boundary markers or sanitization for this untrusted input while possessing capabilities for file writing and command execution. 1. Ingestion points: Reads sprint-status.yaml, *epic*.md, *architecture*.md, *gdd*.md, *prd*.md, and development story records. 2. Boundary markers: Absent; the skill directly interpolates external content into its analysis steps without delimiters. 3. Capability inventory: The skill can write files (Step 11) and execute shell commands (Step 1 and Step 12). 4. Sanitization: Absent; external content is processed and synthesized without escaping or validation.

gds-retrospective