The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from GitHub pull request review threads, review bodies, and conversation comments via the gh api graphql command. This content is used to determine code modifications and generate replies.
Ingestion points: Fetched via GraphQL query in SKILL.md (Stage 2), including review bodies, comments, and thread content.
Boundary markers: The skill employs a functional marker  to track processed items, but lacks explicit security boundaries or instructions for the agent to ignore malicious directives embedded within the comments.
Capability inventory: The skill can perform git commit, git push, and gh api POST operations. It also executes shell commands identified in project manifests.
Sanitization: No sanitization or escaping of the ingested GitHub content is performed before it is passed to decision-making logic.
[COMMAND_EXECUTION]: The skill relies on the gh (GitHub CLI) and git command-line tools to perform its primary functions, including repository identification, PR listing, branch checkouts, and pushing changes.
[DYNAMIC_EXECUTION]: During the verification stage (Stage 5), the skill identifies and executes test, lint, and typecheck commands retrieved from the project's AGENTS.md or manifest file. This behavior allows for the execution of arbitrary commands defined within the target repository.

answer-pr-reviews