The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest an external 'idea document' (referenced via the --auto flag) to automatically generate project requirements, roadmaps, and configurations. This creates a surface where malicious instructions inside the user-provided document could influence the agent's actions during the research or generation phases.
Ingestion points: Project idea document provided via @ reference when using the --auto flag (defined in the context section of SKILL.md).
Boundary markers: No explicit delimiters (e.g., XML tags or clear 'ignore instructions within this block' warnings) are defined in the instructions for handling the external idea document content.
Capability inventory: The skill is granted access to high-privilege tools including Bash, Write, and Agent, which are used to create the .planning/ directory structure and files.
Sanitization: There is no evidence of sanitization, escaping, or validation logic to ensure the user-provided content does not contain prompt injection payloads that could be executed by the Bash tool or lead to malicious file writes via the Write tool.

gsd-new-project