The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions provide a static, hardcoded Client Secret (1234567890) for the user to configure in the Umbraco backoffice. Hardcoding specific secrets in documentation or instructions rather than using placeholders or environment variables is a high-risk practice.
[INDIRECT_PROMPT_INJECTION]: The skill workflow involves reading existing source code, tool indexes, and test infrastructure to generate new test files, which can lead to the execution of malicious instructions if those files are compromised.
Ingestion points: Reads tool definitions from src/umbraco-api/tools/{collection}/index.ts, test infrastructure from setup.ts, and logic from the tool files themselves.
Boundary markers: The instructions do not specify any boundary markers or instructions for the agent to ignore potentially malicious content within the analyzed files.
Capability inventory: The skill has the ability to write new .ts files to the filesystem and execute shell commands via npm run compile and npm test.
Sanitization: There is no evidence of sanitization or filtering of the ingested code before it is used to influence the output of the integration-test-creator agent.

add-test