feat(seller): real on-disk deliverables + dry-run

[openclaw-consensus-bot]

Fixes seller runtime offerings so they are non-stub and produce concrete on-disk artifacts.

Key changes:

• Add JobContext passed to offering handlers (requirements, ctx)
• Seller runtime writes deliverables under ACP_DELIVERY_ROOT//
  • default: /deliverables/acp-delivery (when repo lives under /skills/)
  • fallback: /deliverables/acp-delivery
• Add offering folders + handlers for currently-registered offerings:
  • openclaw_skill_development
  • smart_contract_security_audit
  • ai_agent_liveness_integration
  • typescript_api_development
  • base_defi_development
  • code_review_and_optimization
    Each writes JOB_SNAPSHOT.json + (INTAKE_REQUEST.md if missing required fields, else REPORT.md)
• Add scripts/dry-run-delivery.ts (+ npm script seller:dry-run) to prove files are written
• Update handler template + seller reference docs to match the delivery pattern

Notes:

• Handlers intentionally do NOT claim audits/reviews were performed unless tool output is included.

[openclaw-consensus-bot]

Update pushed for seller delivery hardening:\n\n- now supports , cleans job dirs before each run, validates expected deliverable type per mode, and checks delivery root consistency.\n- handler template no longer scaffolds TODO/stub report text; it scaffolds intake+report flow with compliance note.\n- template snippet updated to match runtime behavior and .\n\nValidation run:\n- v22.22.0 => v22.22.0\n- 10.9.4 => 10.9.4\n-

virtuals-protocol-acp@0.2.0 seller:run
tsx bin/acp.ts serve start --help

acp serve — Manage the seller runtime process

Subcommands:
start Start the seller runtime (listens for incoming jobs)
stop Stop the seller runtime
status Show whether the seller is running
logs Show recent seller logs (last 50 lines)
logs --follow Tail seller logs in real time (Ctrl+C to stop) OK\n-

virtuals-protocol-acp@0.2.0 seller:dry-run
tsx scripts/dry-run-delivery.ts --both

[dry-run] offering=ai_agent_liveness_integration jobId=999000000 mode=written status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000000
[dry-run] offering=ai_agent_liveness_integration jobId=999000001 mode=needs_info status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000001
[dry-run] offering=base_defi_development jobId=999000002 mode=written status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000002
[dry-run] offering=base_defi_development jobId=999000003 mode=needs_info status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000003
[dry-run] offering=code_review_and_optimization jobId=999000004 mode=written status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000004
[dry-run] offering=code_review_and_optimization jobId=999000005 mode=needs_info status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000005
[dry-run] offering=openclaw_skill_development jobId=999000006 mode=written status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000006
[dry-run] offering=openclaw_skill_development jobId=999000007 mode=needs_info status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000007
[dry-run] offering=smart_contract_security_audit jobId=999000008 mode=written status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000008
[dry-run] offering=smart_contract_security_audit jobId=999000009 mode=needs_info status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000009
[dry-run] offering=typescript_api_development jobId=999000010 mode=written status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000010
[dry-run] offering=typescript_api_development jobId=999000011 mode=needs_info status=ok dir=/opt/fundbot/work/workspace-connie/deliverables/acp-delivery/999000011 OK for all offerings (written + needs_info).

[openclaw-consensus-bot]

Follow-up (format-fixed):

Changes in this push:

• scripts/dry-run-delivery.ts: added --both, cleans job dirs before each run (prevents stale-file false positives), validates expected deliverable.type per mode, and verifies resolved delivery root.
• src/commands/sell.ts: refreshed sell init handler scaffold to remove stub/TODO report wording and scaffold a concrete intake/report flow.
• references/seller.md: updated template snippet + dry-run guidance to npm run seller:dry-run -- --both.

Sanity:

• node -v → v22.22.0
• npm -v → 10.9.4
• npm run seller:run -- --help ✅
• npm run seller:dry-run -- --both ✅ (all offerings, both written and needs_info paths)

[openclaw-consensus-bot]

Cycle update (eng-openclaw-acp):

• Added file-linked deliverable payload helpers in seller runtime (reportPath/intakePath, reportUri/intakeUri, fileRefs array).
• Updated all built-in offering handlers plus acp sell init scaffold template to return linked deliverables while still writing canonical files under deliveryRoot/jobId.
• Hardened seller dry-run to assert pointer fields in deliverable payloads (not just file existence/type).
• Docs updated in README and references/seller.md to include dry-run verification and linked-deliverable contract.

Sanity run on this branch:

• node -v => v22.22.0
• npm -v => 10.9.4
• npm run seller:run + seller:check + seller:stop => OK
• npm run seller:dry-run -- --both => OK across all offerings

Commit: e7352f0

[openclaw-consensus-bot]

Cycle update (eng-openclaw-acp):\n\nNew commit: d18d18f\n\nWhat changed:\n- Reworked all 6 built-in offering handlers to produce concrete, offering-specific delivery artifacts (not placeholder report text).\n- Each successful job now writes:\n - JOB_SNAPSHOT.json\n - REPORT.md\n - offering-specific plan/spec artifacts (e.g. AUDIT_PLAN.json, DEFI_BUILD_PLAN.json, API_BUILD_PLAN.json, etc.)\n- Intake path remains strict: missing required fields -> INTAKE_REQUEST.md + structured needs_info deliverable.\n- Updated docs:\n - README selling flow now clarifies dry-run verifies REPORT/INTAKE plus offering-specific artifacts.\n - references/seller.md template now demonstrates writing DELIVERY_PLAN.json + concrete report content.\n\nSanity run:\n- node -v => v22.22.0\n- npm -v => 10.9.4\n- npm run seller:run => OK\n- npm run seller:check => OK\n- npm run seller:stop => OK\n- npm run seller:dry-run -- --both => OK (all offerings, written + needs_info)\n

[openclaw-consensus-bot]

Follow-up hardening:\n\n- Added guard in scripts/dry-run-delivery.ts: written-mode jobs must now produce at least one offering-specific artifact beyond JOB_SNAPSHOT.json + REPORT.md.\n- This enforces non-stub behavior in CI/smoke flows and prevents regressions back to placeholder-only delivery.\n\nCommit: f631366\nValidation:\n- node -v => v22.22.0\n- npm -v => 10.9.4\n- npm run seller:run / seller:check / seller:stop => OK\n- npm run seller:dry-run -- --both => OK\n

[openclaw-consensus-bot]

Cycle update (eng-openclaw-acp):

New commit: 8d96469

What changed:

• Hardened sell create validation: handlers.ts now requires executeJob to accept two params (requirements, ctx), not just be exported.
• Updated acp sell init scaffold to produce concrete written-mode artifacts by default:
  • JOB_SNAPSHOT.json
  • DELIVERY_PLAN.json
  • REPORT.md
• Kept missing-requirements intake path (INTAKE_REQUEST.md + needs_info deliverable).
• Fixed scaffold escaping for the intake JSON example so generated handlers compile cleanly.

Validation:

• node -v => v22.22.0
• npm -v => 10.9.4
• npm run seller:run / seller:check / seller:stop => OK
• npm run seller:dry-run -- --both => OK (all built-in offerings)
• Scaffold smoke test: acp sell init tmp_delivery_contract; npm run seller:dry-run -- tmp_delivery_contract 999002000 --both => OK; temp offering removed.

[openclaw-consensus-bot]

Cycle update (eng-openclaw-acp):

New commit: c2f3036

What changed:

• Hardened seller negotiation memo parsing so offerings resolve across schema variants:
  • offering name from name | offeringName | offering
  • requirements from requirement | requirements | serviceRequirements
  • fallback for inline requirement fields if wrappers are absent
• This prevents false needs_info responses when buyer requirements are present under alternate payload keys.

Sanity:

• node -v => v22.22.0
• npm -v => 10.9.4
• npm run seller:run -- --help => OK
• npm run seller:dry-run -- --both => OK (all offerings, written + needs_info paths; files written under /opt/fundbot/work/workspace-connie/deliverables/acp-delivery//)

[openclaw-consensus-bot]

Cycle update (eng-openclaw-acp):

New commit: 457a915

What changed:

• Hardened scripts/dry-run-delivery.ts to assert clear intake prompts in INTAKE_REQUEST.md (must label intake, list missing fields, and instruct buyer to re-submit with requirements).
• Added schema check so needs_info deliverables must include required fields from offering.json.requirement.required.
• Added REPORT.md quality assertions in dry-run (REPORT heading, artifact section, jobId traceability).
• Added npm test alias to run seller:dry-run -- --both; updated README + references/seller.md docs.

Sanity:

• node -v => v22.22.0
• npm -v => 10.9.4
• npm test => OK (all offerings pass written + needs_info modes; artifacts written to /opt/fundbot/work/workspace-connie/deliverables/acp-delivery//)

[openclaw-consensus-bot]

Update from latest engineering cycle:\n- typescript_api_development now writes a concrete ENDPOINT_CONTRACT.json (inferred endpoint set + auth strategy) alongside API_BUILD_PLAN.json, ENDPOINT_DRAFT.md, and REPORT.md.\n- Removed placeholder/TODO contract text from generated deliverables.\n- seller dry-run now enforces a no-stub-marker rule (TODO/TBD/stub/placeholder) across written artifacts.\n\nValidation run:\n- npm test (seller:dry-run -- --both) ✅