The Agent Skills Directory

[SAFE]: The skill implements a pure-append ledger for tracking build and change history. The analysis shows that all operations are restricted to a local repository configuration directory (.plan/work/change-ledger.jsonl) and use standard Python file handling for atomic writes.
[COMMAND_EXECUTION]: The skill instructions and scripts use argparse and subprocess wrappers (execute-script.py) to manage git-related commands. These are used for standard repository operations like git rev-parse and git diff through the shared worktree_sha.py module. All command arguments are handled as structured parameters, and no arbitrary shell injection vectors were identified.
[DATA_EXPOSURE]: The ledger file stores metadata about builds (exit codes, plan IDs) and commits (hashes, file paths). It does not access or store sensitive credentials, environment variables, or private keys. Access is restricted to the local working directory.

manage-change-ledger