The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the @glasser-ai/cli package from the public NPM registry to enable its core research capabilities.\n- [COMMAND_EXECUTION]: Employs the Bash tool to execute glasser CLI commands for authentication, searching the endpoint catalog, and running data collection tasks.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external third-party providers which could contain malicious instructions.\n
Ingestion points: Third-party API responses (e.g., Reddit, SERPs) read in SKILL.md (Step 7).\n
Boundary markers: Not explicitly defined for the external provider responses.\n
Capability inventory: Bash execution, file Read, and file Write operations across the workflow.\n
Sanitization: Uses JSON file inputs (-f) for the CLI to prevent shell interpolation of user queries, though retrieved data is not filtered for prompt injection before being returned to the agent context.

glasser-research