The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a modular architecture of shell scripts and Python helpers to manage local state and configurations. This includes install.sh for setup, hook-config.sh for JSON management, and settings-mutator.sh for modifying Claude's internal settings.json file. All operations are confined to the local project and the user's home directory.
[PROMPT_INJECTION]: The context-monitor.sh script is designed to steer agent behavior by injecting instructions into the prompt context when token thresholds are exceeded. For instance, in 'auto-handoff' mode, it provides instructions such as "Stop current work. Call the /handoff skill..." While this technically overrides agent autonomy, it is the primary intended functionality of the skill and is implemented through documented platform mechanisms (the additionalContext envelope).
[EXTERNAL_DOWNLOADS]: The skill documentation refers to the gum tool (from Charm Bracelet) to provide an interactive TUI for hook management. This is a well-known and reputable third-party utility, and the skill does not attempt to download or execute it automatically.

memo-hooks