The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes standard development lifecycle scripts (test, typecheck, lint, and build) using the project's detected package manager (npm, yarn, pnpm, or bun). These commands are based on the configurations found in the project's package.json file.
[PROMPT_INJECTION]: The skill identifies and processes instructions from external markdown files located in the plans/ directory. This creates an indirect prompt injection surface where the agent might follow malicious instructions if the underlying issue files are compromised.
Ingestion points: The agent reads README.md, progress.md, and issue-specific markdown files (plans/<plan>/issues/*.md) to determine its task scope.
Boundary markers: The skill does not implement delimiters or safety warnings to distinguish between system instructions and external data; it explicitly instructs the agent to "Follow the issue file exactly."
Capability inventory: The agent has the capability to modify local project files and execute shell commands through the package manager as part of its standard operation.
Sanitization: No validation or sanitization of the content within the issue files is performed prior to the implementation step.

work-on-issue