The Agent Skills Directory

[SAFE]: Analysis of the skill and the referenced implementation playbook indicates that it is a documentation-centric tool for organizing work. It contains no executable scripts, hardcoded credentials, or network-based operations.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external project files, which creates a potential surface for indirect prompt injection. 1. Ingestion points: spec.md, plan.md, and tracks.md (referenced in resources/implementation-playbook.md). 2. Boundary markers: None are explicitly defined for the processing of these files. 3. Capability inventory: The skill specifies workflows for writing track metadata, implementation plans, and interacting with external Git-based tools. 4. Sanitization: The playbook includes defensive instructions to validate track names and confine file paths to prevent directory traversal or symlink abuse.

track-management