The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted codebase data which serves as a potential surface for indirect prompt injection.
Ingestion points: The agent reads project-specific documentation including CONTEXT.md and Architecture Decision Records (ADRs), and walks the codebase using an Explore sub-agent.
Boundary markers: The instructions do not specify boundary markers or instructions to disregard potentially malicious commands embedded in the processed files.
Capability inventory: The skill has the capability to execute git log commands, spawn parallel sub-agents via the Agent tool, and modify documentation files like CONTEXT.md and ADRs.
Sanitization: Content read from the project repository is used to inform agent suggestions without explicit sanitization or validation logic.
[COMMAND_EXECUTION]: The skill performs local command execution to analyze the development environment.
Evidence: The skill uses git log --oneline to identify codebase hot spots based on commit history.
Context: This execution is used solely for metadata analysis of the repository and aligns with the tool's primary purpose of codebase architectural review.

improve-codebase-architecture