The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from pull request diffs and details, creating a surface for embedded instructions to influence the agent's behavior during the review.\n
Ingestion points: The skill uses gh pr view and gh pr diff to bring external content into the context.\n
Boundary markers: Although the skill includes a security warning, it lacks structural delimiters (e.g., XML tags or clear markers) to separate the untrusted diff content from the agent's instructions.\n
Capability inventory: The agent has access to powerful tools including the GitHub CLI (gh), git, and terraform.\n
Sanitization: No explicit sanitization or filtering of the pull request data is performed before processing.\n- [COMMAND_EXECUTION]: The skill uses shell commands to interact with the repository and perform automated checks.\n
Evidence: Usage of gh, git, terraform, and jq for fetching data and managing integration worktrees.\n- [DYNAMIC_EXECUTION]: The workflow requires running terraform plan on code submitted by external contributors.\n
Evidence: Maliciously crafted Terraform files (e.g., using external data sources or malicious providers) can trigger local command execution during the plan phase.

review-pr