The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub (issue titles, PR statuses) and local git metadata (branch names) which could potentially be used to influence the agent's behavior.
Ingestion points: The skill reads external data using gh pr list, gh issue view, and orca worktree list (specifically branch names like issue-<n>-<slug>).
Boundary markers: The skill instructions explicitly require a manual preview and a single confirmation step for both batch metadata updates (link mode) and irreversible deletions (clean mode).
Capability inventory: The skill utilizes shell commands to delete local directories (git worktree remove), delete git branches (git branch -d), and stop active terminal processes (orca terminal stop).
Sanitization: While the skill parses issue numbers from branch naming conventions, it does not explicitly describe sanitization of text data (like issue titles or comments) before interpolation or display.
[COMMAND_EXECUTION]: The skill's primary functionality is implemented through the execution of shell commands for Git, GitHub CLI, and the Orca CLI.
Destructive Operations: The clean mode is designed to permanently remove directories and delete git branches. These operations are limited to worktrees verified as merged and require explicit user approval.
System Configuration: The align mode modifies the Orca application's configuration and creates a temporary worktree for path verification purposes, which is immediately removed.

orcakit