| name | reconnect-prompt |
| description | Turn an expired-token or 401 failure into one copy-paste reconnect prompt and log the failed call. Say "the integration broke", "my token expired", "reconnect <tool>", or "I got a 401". Stops the failing action, never retries, never asks for credentials. Routes the failure to the quarantine catch-net so a dead connector does not stay silent. |
| why | A connected tool that quietly stops working is the worst failure in an integrated OS - the next run looks fine until you notice a week of missing data. This turns a silent 401 into one visible prompt and one logged entry. |
| summary | Turn an expired-token or 401 failure into one copy-paste reconnect prompt. |
| enhance | Run it the moment a tool throws a 401 rather than retrying blind - the logged entry is what lets stall detection see the outage instead of mistaking it for a quiet week. |
| allowed-tools | ["Read","Write","Edit","Glob","Grep"] |
| mcp_requirements | [] |
Reconnect Prompt
Runs on: local-writes - creates or edits files in your OS folder; needs an agent with write access.
Handle expired auth for connected tools without silent retries or automatic re-auth. The skill gives you one clear prompt to paste into the relevant tool's reconnect flow and records the failed call so it does not disappear.
Pre-flight
- If
core/identity.md does not exist, stop with: Founder OS not set up here. Run /founder-os:setup first.
- Read
stack.json to resolve which placeholder failed ({knowledge_base}, {email_platform}, {calendar}, {automation_platform}, {crm}, {file_storage}, {meeting_notes}, {prospecting_db}, {video_tool}, {booking}, {primary_channel}, or another key in that file). If stack.json is missing, proceed with the placeholder name the calling skill passed.
When to invoke
Invoke when an integration-backed action returns any of:
- HTTP 401.
- Expired token.
- Invalid grant.
- Revoked consent.
- Missing refresh token.
- Auth scope no longer valid.
Protocol
- Receive the auth failure from the calling skill: source skill, stack key, action attempted, and a one-line error summary.
- Stop the calling action. Do not retry.
- Produce one reconnect line you can paste into the tool's reconnect UI.
- Tell the operator which
stack.json key failed.
- Log the failure to the quarantine catch-net (see below).
- Return control to the calling skill only after the operator confirms the tool is reconnected.
This skill never asks for credentials and never stores tokens.
Where the failure is logged
The public OS has a silent-failure catch-net at system/quarantine.md (template fallback templates/system/quarantine.md). Append one Status: ACTIVE entry there using the entry format documented in that file: timestamp, source (reconnect-prompt), trigger (the action attempted), error (the summary), context (the stack key). The SessionStart brief counts ACTIVE entries, so the dead connector surfaces at the next session.
If neither nor its template exists on this install, degrade gracefully: produce the reconnect prompt and append a one-line note to instead. Do not invent a quarantine file - one line in the log is enough to keep the failure visible.
