The Agent Skills Directory

CREDENTIALS_UNSAFE (HIGH): The README explicitly lists default administrative credentials (admin@windmill.dev / changeme) for accessing the application UI.
EXTERNAL_DOWNLOADS (MEDIUM): The quick start guide instructs users to install the windmill-cli package globally via npm, which involves downloading and executing code from an external registry without cryptographic verification.
COMMAND_EXECUTION (MEDIUM): The skill's primary function is the orchestration and execution of arbitrary scripts (Python, TypeScript, Bash, Go). Documentation demonstrates using these scripts to perform network operations and system-level tasks.
DATA_EXFILTRATION (MEDIUM): Code snippets illustrate how to retrieve sensitive credentials (API keys, Postgres DB strings) from a central resource manager and use them in external network requests via curl and requests.
PROMPT_INJECTION (LOW): The skill facilitates Indirect Prompt Injection (Category 8) by ingesting untrusted data from APIs and user parameters into execution modules without visible sanitization.
Ingestion points: Untrusted data enters through the api_url parameter in Python scripts and the input string in TypeScript modules.
Boundary markers: There are no delimiters or 'ignore' instructions present to prevent the execution of instructions embedded within the processed data.
Capability inventory: The scripts have access to network tools (curl, requests), resource management secrets (wmill.get_resource), and general shell execution.
Sanitization: No input validation, escaping, or schema enforcement is shown in the examples.

windmill