The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted data from the local repository via the git diff --staged command to generate commit messages. This creates an attack surface for indirect prompt injection, where malicious instructions embedded in staged file changes could influence the agent's behavior or the content of the generated commit.
Ingestion points: The output of git diff --staged is injected into the agent context via dynamic context execution in SKILL.md.
Boundary markers: The diff output is not enclosed in delimiters or accompanied by instructions to ignore embedded commands.
Capability inventory: The skill utilizes the Bash tool to execute repository commands such as git commit and git show.
Sanitization: There is no evidence of sanitization or filtering of the diff output before it is processed by the model.
[COMMAND_EXECUTION]: The skill executes multiple shell commands using the Bash tool. This includes dynamic context commands (git status, git diff) run at skill load time and operational commands (git commit, git show) triggered by user action. The allowed-tools frontmatter configuration restricts the Bash tool scope, which may impact the execution of commands not explicitly permitted by the platform's security policy.

commit