Sigil: Workbench Up

• the user asks to run, reopen, or demo the local workbench,
• the workbench must be available for a ResonantOS or augmentor demonstration,
• a manual polling request needs to be made available without revealing the task before claim,
• the bridge needs a quick L0 contract smoke before real work starts.

Do not use this sigil for remote deployment, browser-extension control, model-provider integration, or real skill execution attestation.

• repository root, resolved from the current working directory,
• ide-extension project at projects/ide-extension,
• desired mode: --start, --check, --available, or --smoke,
• optional task text,
• optional requested skill slug,
• optional claimant identity for smoke mode.

1. Resolve the repository root from the current working directory.
2. Confirm projects/ide-extension/package.json exists.
3. Confirm its start script is node src/server.mjs or an equivalent local server command.
4. Build the base URL as http://127.0.0.1:<port>, defaulting to port 8765.
5. Build the workbench URL as <base-url>/demo/l0.

Step 2 - Check Readiness

1. Request the workbench URL with a short timeout.
2. If it returns HTTP 200, report Server: running.
3. If the user selected --check and the route is not available, return BLOCK with the failed URL and do not start anything.
4. If the route is not available and the mode allows startup, run npm run start from projects/ide-extension as a long-running local server command.
5. Poll the workbench URL until it returns HTTP 200 or the startup timeout expires.
6. If startup fails, return BLOCK with the command, port, and latest connection error.

Step 3 - Prepare Availability

Run this step only for --available and --smoke.

1. Prepare one envelope with POST /api/envelopes/prepare.
2. Use the supplied task text or Workbench bridge smoke.
3. Use the supplied requested skill or decision-gate.
4. Preserve the approval boundary: the prepared request must still be approved before it is ready.
5. Approve the prepared request with POST /api/approvals/<approval_id>/approve.
6. Mark the approved envelope ready with POST /api/bridge/requests/<envelope_id>/ready.
7. Confirm GET /api/bridge/requests/ready returns one opaque request_handle.
8. For --available, stop here and report the envelope ID, approval ID, and ready handle. Do not claim the request.

Step 4 - Run Optional Smoke

Run this step only for --smoke.

1. Claim the ready request with POST /api/bridge/requests/<request_handle>/claim.
2. Bind the claim to agent_ref, defaulting to openrouter.local-session.
3. Use a caller-generated session_ref when one is not provided.
4. Confirm the successful claim reveals the task, requested skill, grants, and context.
5. Submit one synthetic result with POST /api/bridge/claims/<claim_id>/result.
6. Use status: "passed" and result_type: "skill.execution" only for the local smoke, and state that this does not attest real skill execution.
7. Confirm /api/state shows the envelope in a terminal passed state and the ready handle cleared.

Step 5 - Close Honestly

1. Report the workbench URL prominently.
2. State which mode ran.
3. State whether the server was already running or started during the run.
4. State whether an available request was created.
5. State whether a claim/result smoke was run.
6. If any step failed, preserve the exact failing endpoint or command.

• mode,
• server state,
• workbench URL,
• prepared envelope ID when present,
• claim ID and agent_ref when smoke mode claims a request,
• validation commands or endpoints checked,
• proof boundary and any overclaim risk.

## Workbench Up Result
- Mode:
- Server:
- URL:
- Availability:
- Prepared envelope:
- Claimant:
- Receipt:
- Validation:
- Proof boundary:
- Follow-up: