The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation requires the installation of the flowlines package and its instrumentation extras (e.g., flowlines[openai], flowlines[all]) from PyPI. These are legitimate resources owned by the skill author (flowlines-ai).
[DATA_EXFILTRATION]: The SDK is designed to capture LLM telemetry, including requests, responses, and metadata, and export them to the vendor's infrastructure at ingest.flowlines.ai and api.flowlines.ai. This is the primary function of the tool.
[PROMPT_INJECTION]: The presence of get_memory() and aget_memory() functions creates a surface for indirect prompt injection by retrieving data from external storage that might be used in future prompts.
Ingestion points: get_memory() and aget_memory() in SKILL.md fetch JSON data from the Flowlines API.
Boundary markers: No delimiters or instructions to ignore embedded content are provided in the integration examples.
Capability inventory: The skill provides monitoring and memory retrieval functions but does not contain dangerous system-level capabilities like file writing or arbitrary code execution.
Sanitization: The skill returns raw JSON data from its memory retrieval functions without explicit sanitization instructions.
[COMMAND_EXECUTION]: The documentation includes a curl example for verifying trace ingestion that uses unencrypted HTTP (http://api.flowlines.ai). This represents a best-practice violation as it could expose the API key in the x-flowlines-api-key header to interception.

integrate-flowlines-sdk-python