The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes local shell commands including git and the GitHub CLI (gh) to extract repository metadata, diffs, logs, and pull request information. These commands are integral to the skill's primary function of generating documentation.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources such as commit messages, pull request descriptions, and source code diffs. An adversary who can contribute to the repository or open a pull request could potentially embed instructions within these fields to manipulate the agent's behavior during the generation of the walkthrough artifact.
Ingestion points: Data is ingested through git diff, git log, gh pr diff, and gh pr view (specifically the PR body).
Boundary markers: There are no explicit instructions to use delimiters or boundary markers to isolate the retrieved external data from the agent's internal instructions.
Capability inventory: The skill possesses the ability to execute shell commands (git, gh), write to the local file system (walkthrough.html), and publish content via the platform's Artifact tool.
Sanitization: The instructions recommend escaping HTML special characters (&, <, >) in the final output to ensure correct formatting and prevent HTML injection in the browser, but this does not mitigate potential prompt injection during the agent's processing phase.

diff-explainer