The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of system commands and subprocesses, including git, gh (GitHub CLI), glab (GitLab CLI), and jq to manage repository state and interact with remote platforms.
[DYNAMIC_EXECUTION]: The common.sh utility script uses eval within the load_kv_output function to import variables generated by other scripts. While printf %q is used to escape values, the use of eval on dynamically constructed strings is a risky pattern.
[DYNAMIC_EXECUTION]: The normalize_path function in common.sh performs runtime script generation and execution by passing a heredoc to python3 for path normalization tasks.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from external sources and acts upon it.
Ingestion points: Fetches pull request comments and discussion threads via scripts/fetch_review_comments.sh; parses repository-level policy files (AGENTS.md and CLAUDE.md) in scripts/repo_policy.sh.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands were found in the processing logic.
Capability inventory: The skill possesses capabilities to modify local git state, manage worktrees, and post comments to remote APIs using scripts/worktree_sync.sh and scripts/post_review_comment.sh.
Sanitization: Employs jq for JSON data and printf %q for shell variable escaping, though repository policy parsing relies on regular expressions which may be bypassed.
[EXTERNAL_DOWNLOADS]: The skill interacts with well-known services (GitHub and GitLab) to fetch pull request metadata and comments. These operations use official CLI tools and require existing user authentication.

review-pr