The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple Git commands, including history-altering operations like git reset --soft and git push --force-with-lease. These commands are standard and necessary for the skill's primary function of squashing commits.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from the repository environment to perform semantic analysis.
Ingestion points: The agent reads untrusted data from commit logs (git log) and code changes (git diff) to determine the intent and content of the squashed commit.
Boundary markers: The instructions lack explicit delimiters or safety instructions to distinguish between the agent's internal logic and potentially malicious instructions embedded within commit messages or code diffs.
Capability inventory: The skill possesses the capability to execute shell commands (git), write files (temporary commit message files), and interact with remote servers via Git push.
Sanitization: There is no evidence of sanitization or filtering applied to the commit logs or diff content before the agent processes them to derive a commit message.

git-squash