The Agent Skills Directory

Command Execution: The skill instructs the agent to utilize several shell-based development tools to validate code correctness and project history. These tools are used to gather objective evidence for code reviews.
Evidence: Usage of npx tsc, npx eslint, git log, grep, and npx oxlint as described in SKILL.md and common-patterns.md.
Indirect Prompt Injection Surface: The skill is designed to ingest and review user-provided code files and git metadata. This processing step introduces a surface where malicious instructions could be embedded within code comments or documentation blocks to influence agent behavior.
Ingestion points: Processes TypeScript/JavaScript source files, wrangler configuration files (.jsonc, .toml), and git history via git log (found in SKILL.md).
Boundary markers: The skill relies on external reference sources like config-schema.json and workers-types to establish validity, though it does not explicitly define injection-specific delimiters for the code under review.
Capability inventory: The agent is instructed to execute shell-based commands including npx tsc, npx eslint, npx oxlint, git, and grep (found in SKILL.md and common-patterns.md).
Sanitization: The instructions emphasize strict validation against type definitions and JSON schemas to ensure that the code and configuration conform to expected platform patterns.

workers-code-review