The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of external OpenAPI specification files through an AI-powered suggestion tool, which represents a potential ingestion point for untrusted data.
Ingestion points: OpenAPI spec file path provided via the -s argument in SKILL.md.
Boundary markers: None explicitly defined in the instructions to delimit or warn about potentially malicious content within the specification file.
Capability inventory: The skill uses the speakeasy CLI to perform network requests (fetching suggestions) and can write the output to a local file using the -o flag.
Sanitization: No explicit sanitization or validation of the input spec file is performed within the scope of these instructions.
[COMMAND_EXECUTION]: The skill defines several commands using the speakeasy CLI tool, which is the primary intended functionality of the skill author.
Evidence: Usage of speakeasy suggest operation-ids, speakeasy suggest error-types, and speakeasy run are documented as standard operations for SDK improvement.
[CREDENTIALS_UNSAFE]: The skill provides instructions for handling authentication which align with security best practices.
Evidence: It correctly advises users to set the SPEAKEASY_API_KEY as an environment variable in non-interactive environments, rather than hardcoding credentials.

improve-sdk-naming