The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: A URL referenced in the workflow documentation (https://product.md/) was flagged as malicious and blacklisted by automated security scans.
[REMOTE_CODE_EXECUTION]: The integration guidance in references/integrations.md instructs the agent to download and execute the Sentry MCP server package using pnpm dlx. This is documented neutrally as a feature of a well-known service but involves runtime execution of external code.
[COMMAND_EXECUTION]: The skill defines a large number of complex shell commands within its configuration templates (templates/hard-eng.dart.json, templates/hard-eng.javascript.json, templates/hard-eng.python.json) for tasks like code formatting, linting, and security auditing, including the use of shell scripts (sh -c), xargs, and piping.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external project files like PRODUCT.md and DESIGN.md, as well as external service APIs. Since it lacks explicit boundary markers or sanitization for this content, it presents an attack surface for indirect prompt injection that could influence the agent's behavior during task execution.