The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script scripts/calculate_experiment_plan.py to perform statistical calculations. The script uses standard libraries (json, math, sys), implements its own mathematical functions (Acklam's approximation) to avoid external dependencies, and includes robust validation for all input parameters.\n- [DATA_EXPOSURE]: The skill interacts with experiment metadata and metrics via the ae-cli tool. It adheres to a strict project gate policy, requiring project resolution before any reads or writes. SDK integration templates use placeholders (e.g., YOUR_SERVER_URL, APP_ID) to ensure no sensitive information is hardcoded.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests platform metadata (events, properties, metrics) using ae-cli commands. While this introduces an external data ingestion point, the skill contains explicit instructions to treat this data as internal working context and strictly validates it against platform-provided schemas, mitigating the risk of instructions embedded in metadata being obeyed.\n- [REMOTE_CODE_EXECUTION]: The skill does not perform any remote code downloads or execution. Documentation references to external SDK packages (Maven, CocoaPods, NPM) are for user-side implementation guidance only and are not invoked by the agent during runtime.

ae-experiment-design