The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from local repository files and diffs to determine commit scope and generate commit messages.
Ingestion points: File content inspection and git diff outputs in SKILL.md (Workflow Step 1 and Step 4).
Boundary markers: The skill uses single-quoted HEREDOCs ('EOF') when executing git commit to prevent the shell from interpreting the content of the commit message.
Capability inventory: The skill has the ability to stage files (git add), create commits (git commit), and send code to remote servers (git push).
Sanitization: The references/ignore-gate.md file and SKILL.md Step 2 provide explicit instructions to identify and block secrets (e.g., .env files, private keys) from being staged or committed.
[COMMAND_EXECUTION]: The skill executes various Git and shell commands to perform its duties. These operations are directly related to its stated purpose of automating repository updates and include safeguards such as dry-run preflights and state verification.

commit-push