The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external API specifications (OpenAPI, GraphQL, Smithy) sourced from third-party endpoints. These untrusted descriptions represent an attack surface where malicious content could attempt to influence the agent's behavior during the SDK conversion and generation process.\n
Ingestion points: Remote data enters the context through fetch-specs.ts scripts located in stacks/distilled-submodules/spec-repos/<pkg>/.\n
Boundary markers: The documentation recommends structural validation (e.g., checking for openapi or paths fields) but lacks explicit security-focused sanitization or instruction-ignoring delimiters.\n
Capability inventory: The skill uses pnpm and bun to execute scripts that perform file operations, code generation, and project maintenance.\n
Sanitization: No specific prompt-injection sanitization or escaping of the external content is described for the generation pipeline.\n- [DYNAMIC_EXECUTION]: The workflow involves creating and running local scripts (fetch-specs.ts) to manage specification retrieval. Additionally, the skill uses the DISTILLED_SPECS_LOCAL environment variable to dynamically alter file path resolution for specifications at runtime.\n- [EXTERNAL_DOWNLOADS]: The pipeline fetches API specifications and documentation from remote sources, including raw.githubusercontent.com and other external provider endpoints.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute various shell commands and project scripts, such as pnpm generate, pnpm convert, pnpm specs:local, and bun install, to automate the SDK lifecycle.

distilled-sdk