The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a highly invasive shell script scripts/setup-worktrees.sh. This script converts the repository into a bare repo layout by cloning to a .bare folder, removing all files and hidden directories from the project root, and then creating worktree pointers. While it includes checks for a clean git status, these destructive file operations pose a risk to user data if misused.
[COMMAND_EXECUTION]: The skill uses node scripts/kata-lib.cjs to perform various configuration tasks, including reading and writing settings to the .planning/config.json file. It also performs git operations like git add and git commit to manage planning artifacts.
[INDIRECT_PROMPT_INJECTION]: The skill reads settings from .planning/config.json via a helper script and interpolates these values directly into the prompt displayed to the user via AskUserQuestion. If the configuration file contains malicious strings, they would be processed and presented by the agent.
Ingestion points: .planning/config.json read via node scripts/kata-lib.cjs in SKILL.md (Process Step 2).
Boundary markers: None; values are interpolated directly into Markdown tables and descriptive text for user review.
Capability inventory: The skill has the capability to execute shell commands (bash, node), write files (fs.writeFileSync in kata-lib.cjs), and modify git repository structure.
Sanitization: There is no evidence of sanitization or escaping of the configuration values before they are rendered in the agent's output display.

kata-configure-settings