The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script scripts/run-fx-review.mjs retrieves API credentials for the vercel-ai-gateway service from the macOS Keychain by executing the /usr/bin/security command. While the script implements redaction logic to prevent these credentials from appearing in logs or artifacts, the retrieval of plain-text secrets is a sensitive operation.- [COMMAND_EXECUTION]: The skill performs extensive command execution across multiple files. The bash script scripts/gate.sh invokes git for diffing and searching repository content, while Node.js scripts in the scripts/ directory execute bun, node, and the fx (Vercel AI Gateway) CLI tool.- [DYNAMIC_EXECUTION]: The test script scripts/test-report-gate.sh utilizes the Function constructor to dynamically execute JavaScript strings. This is used to mutate JSON fixtures for verification purposes within the test environment.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data in the form of git diff output and repository files to determine review logic and triggers. These inputs are interpolated into prompts sent to the AI reviewer in scripts/run-fx-review.mjs, creating a surface for potential indirect prompt injection attacks.

review-gate