The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes bundled shell scripts including prefetch.sh and post_review.sh to perform its workflow. These scripts utilize the GitHub CLI (gh), Git, and the ClickUp CLI to manage data. These operations are essential for the skill's stated purpose of PR review automation.
[PROMPT_INJECTION]: The skill processes untrusted external data which constitutes a surface for indirect prompt injection.
Ingestion points: Untrusted data enters the context through files such as diff.txt, meta.json, existing PR comments (comments.json), and ClickUp ticket summaries (clickup_summary.txt).
Boundary markers: The instructions do not define specific delimiters or "ignore" instructions for the ingested PR content.
Capability inventory: The agent has the ability to execute shell commands, read/write files in the working directory, and post comments to GitHub via the API.
Sanitization: The skill does not implement explicit sanitization of the PR or ticket content, relying on the model's inherent processing logic.
[DYNAMIC_EXECUTION]: The post_review.sh script sources a local configuration file (context.env) that is generated at runtime. The parameters stored in this file are extracted from structured GitHub API responses or validated against restrictive regular expressions, reducing the risk of unauthorized command execution.

review-pr-clickup