The Agent Skills Directory

[SAFE]: The skill's functionality is confined to managing local project state within a specific directory structure (.planning/).
[COMMAND_EXECUTION]: The skill uses git status and git log to determine the current state of the repository and identify in-flight work. These are standard, non-destructive development commands used for state tracking.
[DATA_EXFILTRATION]: The skill does not contain any network-related commands (like curl or wget) or instructions to transmit data externally. All operations are local to the user's project environment.
[PROMPT_INJECTION]: The instructions contain strong behavioral constraints ("Iron Law", "Red Flags") to ensure the AI follows the handoff process correctly, but these are functional guidelines rather than attempts to bypass security or override agent safety protocols.

writing-handoff