The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests and processes untrusted data from external sources.\n
Ingestion points: PR comments are retrieved via the GitHub CLI in scripts/fetch_review_comments.py, and code diffs are read using git in scripts/incremental_review_mapper.py and scripts/build_snapshot.py.\n
Boundary markers: There are no explicit delimiters or instructions to the agent to disregard instructions potentially embedded within comments or patches.\n
Capability inventory: The agent can write files and execute git/gh commands across multiple scripts.\n
Sanitization: No sanitization or filtering of the external text content is performed before it is processed by the agent context.\n- [COMMAND_EXECUTION]: The skill's Python scripts execute system commands using the subprocess module to manage the review lifecycle.\n
Evidence: Found in scripts/fetch_review_comments.py (executing gh), scripts/build_snapshot.py (executing git), and scripts/incremental_review_mapper.py (executing git).\n
Context: These operations are functional for the skill's purpose but involve processing user-provided identifiers like branch names, PR numbers, and commit SHAs.

code-review