The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill employs the !command syntax to execute git log and git diff at load time. These commands are project-specific and benign, serving to provide the necessary context for summarizing branch changes. They do not access sensitive file paths, perform network exfiltration, or incorporate unvalidated user input into the shell execution environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from the repository's git history and diffs.
Ingestion points: Untrusted data enters the context via the output of git commands injected into the SKILL.md file.
Boundary markers: The instructions do not define explicit delimiters or warnings for the agent to ignore instructions that may be embedded within commit messages or diff content.
Capability inventory: The skill's environment is restricted to the specific git commands listed in the allowed-tools frontmatter (git log, git diff, git rev-parse).
Sanitization: No filtering, escaping, or validation is performed on the ingested git output before it is summarized by the agent.

pr-summary