The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local version control commands (git and jj) to inspect changes and record commits.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface due to its ingestion of untrusted data from diffs.
Ingestion points: The workflow reads content from git diff or jj diff commands.
Boundary markers: Absent; the prompt does not use markers to distinguish code data from potential embedded instructions.
Capability inventory: Version control operations (git, jj) are used to read from and write to the local repository.
Sanitization: Absent; no filtering is applied to the retrieved diff content before it is processed by the model.

writing-commit-messages