The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill ingests untrusted data from the project context (files and git history). However, the potential for adversarial influence is mitigated by the incremental presentation (200-300 word sections) and the requirement for human validation before any data is persisted.
[Command Execution] (LOW): The skill utilizes git commands for documentation versioning. These operations are limited to specific documentation paths and follow a human-verified design phase.
[Data Exposure] (LOW): The agent reads local project files to establish context. No network capabilities or exfiltration patterns were identified in the skill logic.

brainstorming