The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill executes shell commands using the exclamation-backtick syntax during initialization to gather repository context. Evidence includes !uv run prek run >/dev/null 2>&1 || git add -u, !git --no-pager status || true, !git --no-pager diff --cached || true, and !git --no-pager diff --cached --stat || true in SKILL.md. These commands fetch status and diffs, and may automatically stage tracked changes (git add -u) if the prek tool fails.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted user input by interpolating $ARGUMENTS into the model prompt. Ingestion points: The $ARGUMENTS variable is placed within <user_notes> tags in SKILL.md. Boundary markers: The input is wrapped in <user_notes> tags, providing structural separation but not complete protection against adversarial instructions. Capability inventory: The skill has the capability to execute git commit and git add commands based on instructions derived from the input. Sanitization: No explicit sanitization or instructions to ignore instructions within the user notes are present.
[COMMAND_EXECUTION]: The skill performs repository management operations using system commands. It is designed to invoke git commit with a generated message and uses uv run to execute linter or pre-commit tools like prek and pre-commit.

commit-staged