The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill provides code templates for instrumenting AI provider calls, creating a surface where the agent-generated code may process untrusted data.
Ingestion points: Code examples in references/openai-tracking.md, references/anthropic-tracking.md, references/gemini-tracking.md, and references/strands-tracking.md demonstrate passing user_prompt or user_input variables directly into model invocation functions.
Boundary markers: The provided templates do not include delimiters or instructions to ignore embedded prompts within the user-supplied data.
Capability inventory: The instrumented code possesses the capability to perform network requests to external AI provider APIs (OpenAI, Anthropic, Gemini, Bedrock).
Sanitization: The skill's instructions and references do not demonstrate or recommend sanitization, validation, or filtering of the untrusted input before processing.- [COMMAND_EXECUTION]: The skill documents how to interact with the LaunchDarkly REST API and external AI services using standard developer tools.
Evidence: references/metrics-api.md and references/langchain-tracking.md provide examples using requests and curl to interact with app.launchdarkly.com. These operations target vendor-owned infrastructure and well-known services, and are documented neutrally.

built-in-metrics