The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/build_hints.py implements a feature that allows the execution of arbitrary shell commands provided through the --llm-ideas-cmd argument. This uses subprocess.run on a command string parsed by shlex.split, enabling the execution of any local binary accessible to the environment.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted market data and news reactions from external files which are then structured for downstream LLM synthesis.\n
Ingestion points: Data enters the system through read_market_summary, read_anomalies, and read_news_reactions in scripts/build_hints.py.\n
Boundary markers: The skill does not implement explicit boundary markers or instructions to ignore embedded commands within the ingested content.\n
Capability inventory: The skill has the capability to write files and execute external subprocesses via the --llm-ideas-cmd hook.\n
Sanitization: No content-based sanitization or filtering is applied to the ingested data values before they are used in hint generation.

edge-hint-extractor