The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates complex tasks by executing shell scripts in sibling skill directories (e.g., memory, assistant, extract-entities) using subprocess.run and subprocess.Popen within the runner.py, storage.py, and report.py modules.
[EXTERNAL_DOWNLOADS]: The runner.py module communicates with a local Lean4 compilation service at http://localhost:8604/compile via HTTP POST requests to verify formal claims.
[SAFE]: The skill incorporates an internal 'Adversarial Reasoning Framework' via a 7-gate decision process. This includes 'Gate 2b: Grounding', which specifically identifies fabricated identifiers and misspelled framework terms to prevent the agent from accepting and reasoning over malicious or hallucinated data.

create-evidence-case