The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a bash script for discovering artifact directories and uses git worktree and cp commands to manage files and checkouts.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8) as it ingests untrusted data from the local file system to determine execution parameters.
Ingestion points: Directory names under docs/plans/ and the content of $ARGUMENTS/repos.md are used to determine branch names, repo paths, and mode of operation.
Boundary markers: None. The skill does not use specific delimiters or instructions to ignore embedded commands within the processed file data.
Capability inventory: The agent can execute shell commands via bash, perform git operations, and copy files with cp.
Sanitization: The skill includes logic to sanitize branch names by replacing slashes with hyphens, providing a degree of safety against ref-path collisions.
Mitigation: The skill implements a mandatory human-in-the-loop checkpoint via AskUserQuestion which displays all derived paths and branch names for user approval before execution.

team-worktree