The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes review notes from .comview/comments.json and treats them as actionable instructions.
Ingestion points: Data is read from the body field of the .comview/comments.json file.
Boundary markers: There are no explicit markers or safety instructions to distinguish between benign review feedback and potentially malicious prompts embedded in the comments.
Capability inventory: The agent is authorized to perform file edits ('make the change') and execute shell commands (git, gh, jj) based on the content of the review notes.
Sanitization: The skill does not specify any sanitization or validation of the comment text before the agent acts upon it.
[COMMAND_EXECUTION]: The skill utilizes several standard command-line interface tools to inspect code diffs.
Evidence: The workflow involves running git diff, git show, gh pr diff, and jj --no-pager diff --git to gather context for the review.

comview