The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes content from .turbo/improvements.md, which serves as an ingestion point for untrusted data. No explicit boundary markers or instructions to ignore embedded directives are used during the read operation. The skill has capabilities including file system modifications (read, write, delete), shell command execution via git, and tool invocation. No sanitization is performed on the file content before it is presented to the user. However, the risk is minimized by the mandatory use of the AskUserQuestion tool for every entry decision, ensuring a human-in-the-loop for all actions.
[COMMAND_EXECUTION]: The instructions require the agent to execute git rev-parse --show-toplevel, git rev-parse --git-dir, and git rev-parse --git-common-dir to correctly identify the repository root and handle linked worktrees for file operations.

triage-improvements