The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the a7 CLI to create, inspect, and update gateway resources. It provides several shell scripts (e.g., in references/recipes/canary.md and references/recipes/blue-green.md) that demonstrate how to automate these operations using the CLI and curl for verification.
[DYNAMIC_EXECUTION]: The skill documents the configuration of the serverless and ext-plugin plugins. These features allow the API7 gateway to execute custom Lua functions (references/plugins/serverless.md) and external binaries written in Go, Java, or Python (references/plugins/ext-plugin.md). These are documented as core extension points for the managed gateway service.
[INDIRECT_PROMPT_INJECTION]: The skill includes instructions for configuring AI gateway features that process untrusted data from external LLM providers.
Ingestion points: Untrusted data enters the agent context when the gateway is configured to proxy requests to AI providers via the ai-proxy plugin (documented in references/plugins/ai-proxy.md).
Boundary markers: The ai-prompt-template and ai-prompt-decorator plugins are used to define fixed structures and system instructions that serve as boundaries for user-supplied prompts (documented in references/plugins/ai-prompt-template.md and references/plugins/ai-prompt-decorator.md).
Capability inventory: The a7 CLI has the capability to modify gateway configuration, while the gateway itself can perform network operations and execute custom logic via other plugins.
Sanitization: The documentation describes using the ai-content-moderation plugin to detect and block toxic or harmful content in AI requests and responses (documented in references/plugins/ai-content-moderation.md).

a7