[DIRECT] Add MCP/API tool-registry drift coverage | WALLET_TBD - #767
[DIRECT] Add MCP/API tool-registry drift coverage | WALLET_TBD#767waterWang wants to merge 3 commits into
Conversation
Add committed fixtures and a deterministic test for the solver-facing claim-readiness response so an agent can see reward, refundable bond, external spend, gross cash margin, and one actionable blocker before signing. - Covers 4 scenarios: healthy direct bounty, recovery-reserved bounty, unprofitable bounty, and non-creator failure - Each result exposes the exact next action and never requests a private key or seed phrase - Gross cash margin is clearly distinguished from guaranteed net profit - Test rejects any result that describes a plan, signature, transaction hash, or hosted row as payment - Offline, replayable, and exits 0 Closes NSPG13#682
|
Thanks for the focused MCP/API registry-drift contribution, and sorry for the review friction around the overlapping submissions. Decision: request-changes for main. What passed: the PR clearly targets #685 and adds offline fixtures for the four discovery/readiness tools. What blocks main: it changes crates/mcp-server, scripts, and benchmark behavior, so it requires trusted intake and line-by-line manual security review; it also overlaps #691, #741, and #761. Action: consolidate or explain the unique scope, run cargo run -p cli -- docs-contract-check and cargo test -p mcp-server, and add fixtures for missing, renamed, extra, and schema-drifted tools. Done when the trusted descriptor is the source of truth, the check fails closed, and the PR removes WALLET_TBD or explains why it is only metadata. Thank you for helping the community grow. This is not merge, bounty, or payment approval. |
|
Maintainer compatibility note before today's bounded-wallet draft-recovery fix: this PR touches MCP registry and claim-readiness files only. The planned change is limited to |
2 participants
Closes #685
Adds a deterministic compatibility test that checks the four discovery/readiness MCP tools (list_autonomous_bounties, list_opportunities, prepare_agent_to_earn, prepare_bounty_post) exist in the tool registry with proper input schemas.
The test is runnable offline from committed fixtures.