You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
| F4 | Owner interacts in real time from a separate Logos app instance via Logos Messaging, no intermediary server |**DONE**|**Recorded on camera** (`docs/lp0008-owner-console.mp4`): a **separate client** holding an owner token (`issue-token --name owner`) reads the agent's status (balance, `per_tx_limit`), sets the spending limit, opens an over-limit task that is **held** (`pending_approvals: 1`), and messages over the owner channel — no intermediary server. Also in `docs/EVIDENCE_LOCAL.md` §3. `basecamp-app/` is the GUI surface over the same channel (assets + build instructions provided). |
136
136
| F5 | Spending threshold: holds above-threshold for owner approval; executes below-threshold autonomously |**DONE**| The threshold gate is verified at the state-machine level (`docs/EVIDENCE_LOCAL.md`): an over-limit task is held as `pending_approval`; `approve_pending` releases it and a below-limit task runs without holding. The released payment settles via the same real-proof wallet primitive as F8 (the integrated agent-wallet hop has the F8 reset-chain caveat). `docs/SECURITY_MODEL.md` covers the gate logic. |
137
137
| F6 | All default skills (storage.*, messaging.*, wallet.*, program.*, agent.*, meta.*) implemented and documented |**DONE**|`scaffold/src/agent_module_impl.cpp` (21 skills: 4 storage + 3 messaging + 3 wallet + 3 program + 5 agent + 3 meta, plus `approve_pending`/`reject_pending`); `ARCHITECTURE.md §7`; `docs/SKILL_INTERFACE.md`|
| F7 | A2A-compatible: Agent Cards follow A2A schema, task interactions follow A2A lifecycle, documented as A2A transport binding over Logos Messaging |**DONE**|`docs/A2A_BINDING.md` (full binding spec); Agent Card is a **SIGNED JSON document** per the spec: A2A `AgentCardSignature` — detached RFC 7515 JWS, EdDSA/Ed25519 (vendored TweetNaCl, canonical source, sha256-pinned), JWK public key in the header, self-verified before publishing (verified live 2026-07-10: `signatures present: 1`, `alg: EdDSA`); AgentSkill entries carry required id/name/description/tags; discover + task lifecycle demonstrated integrated on the live testnet|
139
139
| F8 | Two or more agents discover, execute A2A task lifecycle, transfer LEZ payment autonomously without owner intervention | **DONE** | Both legs demonstrated real and agent-driven. **Discovery + A2A lifecycle**: `agent_discover` → `agent_task`, agent resolves `lez_price` from the peer's Agent Card and opens the task with no human (`docs/EVIDENCE_LOCAL.md` §1a–1c). **Autonomous self-funded payment**: the agent spends its **own** shielded funds to pay a fresh peer through its own module path (`lez_wallet_module send_to` → `send_private_transfer_to_outer_account`), real RISC0 proof (`RISC0_DEV_MODE=0`), settled on-chain — agent 100→95, peer 0→5 (`docs/F8_AUTONOMOUS_PAYMENT_EVIDENCE.md`). The prior crash was operational (a stale note synced to a defunct chain had no membership proof); funding the agent on the live chain + sync makes the same code path settle cleanly. Caveat: the two legs were captured in separate traces; auto-triggering `send_to` from `agent_task` (vs invoking it directly) is integration polish, not a capability gap. |
140
140
| F9 | At least 3 illustrative use cases demonstrated end-to-end on LEZ testnet |**DONE**|`tests/demo-usecases-testnet.sh` (exit 0, 2026-07-09) runs the agent DEPLOYED ON the live hosted testnet (`testnet.lez.logos.co`, v0.2.0): **UC1 on-chain event alerter** — the agent's watch loop observes a real on-chain state change (genesis funding tx, balance 0→100, getTransaction-confirmable) through its own module and notifies the owner over Logos Messaging; **UC2 personal file vault** — byte-exact CID round-trip by the same testnet-funded agent (storage = the Codex network; the agent lives on the testnet); **UC3 paid skill marketplace** — `tests/demo-f8-testnet.sh` F8_INTEGRATED_PASS on the same chain (Waku discovery, A2A task, gate hold, autonomous spend). Additional local end-to-end evidence: `docs/F8_LINUX_FULL_EVIDENCE.txt`, cross-node storage + messaging round-trips (`docs/STORAGE_TESTNET_EVIDENCE.md`, `docs/MESSAGING_TESTNET_EVIDENCE.md`). |
141
141
| F10 | Three separate agents deployed on LEZ testnet — one per skill category (Storage, Messaging, Blockchain) |**DONE**| Three shielded agents (Blockchain/Storage/Messaging), one per category, funded from genesis on the **live** hosted testnet (LEZ `v0.2.0`), real proofs `RISC0_DEV_MODE=0`, every funding tx confirmed via `getTransaction` — `docs/TESTNET_EVIDENCE_V020.md`. Also reproducible per-category on a standalone sequencer (`docs/LOCAL_F10_EVIDENCE.md`). Compatibility gate: wallet built from `v0.2.0` passes `check-health` against the hosted chain. |
0 commit comments