The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted project data, creating a potential surface for indirect prompt injection if project files or commit history contain adversarial instructions.
Ingestion points: The skill reads project files, documentation, and recent commit history to understand context (SKILL.md).
Boundary markers: There are no explicit instructions or delimiters provided to the agent to differentiate project data from its own core instructions.
Capability inventory: The skill utilizes Read, Write, and Bash tools to document designs and interact with the repository.
Sanitization: There is no evidence of sanitization or filtering of the project data before it is processed by the agent.

brainstorming