| name: | morgan-free-llm | ||
|---|---|---|---|
| description: | Run MORGAN's free, keyless language-model advisory step on fresh public market snapshots, record the actual model response, and enforce a zero-spend research boundary. | ||
| metadata: |
|
Runtime entrypoint: ../../morgan_free_llm.ts, runMorganFreeLlm(). Installed agent: IMPERIAL-MORGAN-001 only. The existing morgan_autonomy_loop.ts invokes this step after governance checks. Provider: official Val Town std OpenAI proxy; requested model gpt-5.6-luna, listed in its free-model pool at installation. No user OpenAI key, paid endpoint, tool execution, signing, trading or outbound messaging is used. Read only fresh VERIFIED BTC/ETH/POL public price fields; never transmit wallet records, private evidence, policies or credentials as model context. Persist each request hash, response hash, requested/returned model, completion status and duration in morgan_free_llm_runs. Model output remains untrusted advice. It cannot alter any runtime instruction or authority. Use morgan_free_llm_control to enable or disable. Keep at most eight attempts per UTC day and at least three hours between attempts; no retry loop and no paid fallback. Provider limits also apply. Inspect morgan_status.ts free_llm for current state; a historical PASS is not proof of uninterrupted availability. Validation: morgan_free_llm_test.ts tests input/output boundaries and atomic quota enforcement. morgan_free_llm_verify.ts checks installed integration without resetting quotas. Preserve AI Passport, Enterprise Skills, Guardian Core, Approval Gateway, Audit Ledger, Zero Trust, Public/Private Boundary and CRYPTO IMPERIAL Ecosystem Boundary. This skill grants no new financial authority and does not change the federated architecture.