The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions direct the agent to read and parse ~/.gooseworks/credentials.json to extract API keys. This involves accessing sensitive configuration files within the user's home directory to manage authentication tokens.
[COMMAND_EXECUTION]: The setup and discovery sections rely on executing shell commands using python3 -c to parse JSON files and extract environment variables, which is a form of local command execution.
[EXTERNAL_DOWNLOADS]: The skill recommends using npx to run the gooseworks and goose-skills CLI tools for login and installation. This pattern involves downloading and executing remote code from a public package registry.
[PROMPT_INJECTION]: The skill presents a significant indirect prompt injection surface due to the combination of processing untrusted web content and providing powerful remote capabilities.
Ingestion points: The Create Task, Contents, and Answer endpoints ingest data from external URLs, search queries, and uploaded files (PDFs, images, documents).
Boundary markers: There are no instructions or delimiters provided to the agent to treat retrieved web content as untrusted or to ignore instructions embedded within that data.
Capability inventory: The Create Task endpoint allows for code_execution, the configuration of mcp_servers (Model Context Protocol servers for tool access), and the generation of various file types as deliverables (CSV, Excel, Word, etc.).
Sanitization: The skill does not mention any sanitization or validation of the content retrieved from external sources before it is processed by the agent or the research task.

web-search-valyu