The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill employs the !command`` syntax within SKILL.md to automatically fetch and inject the current repository state into the agent's context.
Evidence: The following commands are executed at load time: git status, git diff HEAD, git branch --show-current, and git log --oneline -10.
Assessment: These operations are consistent with the skill's primary purpose of managing git commits. They provide the agent with the necessary visibility into staged changes and recent history to perform its task correctly. No exfiltration or sensitive file access was observed.
[INDIRECT_PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection because it reads and processes external data (git history, diffs, and status) that could contain malicious instructions.
Ingestion points: Untrusted content from the repository (e.g., commit messages or code changes) enters the context via the git status, git diff, and git log commands in SKILL.md.
Boundary markers: There are no boundary markers or delimiters used to isolate the git output, nor are there instructions for the agent to ignore embedded commands within that data.
Capability inventory: The agent is authorized to perform git add and git commit operations based on the ingested context.
Sanitization: No sanitization, escaping, or filtering is applied to the output of the git commands before they are interpolated into the prompt.

commit