UAPI Agent Continuation Automation Runbook
Hourly heartbeat, diary freshness, daily rollup, stale-work, and continuation packet rules.
Promotion Metadata
- Source feedback id:
82 - Source feedback ids:
82 - Promotion request id:
17 - Feedback category:
consider - Feedback source:
ai-lane - Feedback created at:
2026-06-22T16:31:04.769241+00:00 - Target slug:
[REDACTED:high-entropy] - Review state:
review - Reviewer note: AI lane auto-promoted redacted feedback into internal catalog context.
Source Feedback Body
Feedback 82
- Product:
uapi - Category:
consider - Related slug:
[REDACTED:high-entropy] - Source:
ai-lane - Created at:
2026-06-22T16:31:04.769241+00:00
title: UAPI Agent Continuation Automation Runbook project: uapi kind: runbook state: review tags: [docs, runbook, uapi, automation, continuation, diary] source_plan_title: UAPI Integrated Operating Plan v2
UAPI Agent Continuation Automation Runbook
Purpose: define hourly active-work checks, daily rollups, stale-work detection, missing-audit checks, and continuation packets for Codex, Grok, and GLM 5.2. Automation monitors and reports; it does not create hidden autonomous infrastructure work.
Access contract: Endpoint: http://10.0.0.121:8080. Token: set UAI_EXPLORER_AI_API_KEY out of band. Header uses Bearer token from that variable. Never print secrets. Fetch master program control and this runbook before automation changes.
Active-work heartbeat record: - agent: codex, grok, glm52, or automation - runbook_title - session_id or thread id when available - current_task - last_seen_utc - last_audit_event_id - source_context_titles_or_slugs - status: active, blocked, waiting-review, stale, complete - next_action - blocker - no_secret_confirmed: true
Hourly checks: 1. GET /ready and /api/ai/bootstrap. 2. Fetch active diary/context entries and required runbooks. 3. Flag stale if active work has no fresh diary within 90 minutes or has missing audit ids. 4. Create a continuation packet only when useful: exact next step, contexts to fetch, audits, blocker, and safety boundary. 5. Do not mutate lab devices, run hardware jobs, or promote claims without the assigned agent evidence.
Daily rollup: - summarize each agent stream, completed gates, open contradictions, fresh audit ids, stale items, and next 24-hour priorities. - confirm all live runbooks still fetch through the AI context endpoint with raw_available_to_client=false.
Verifier checks: - Explorer publication: feedback intake is not success; context fetch 200 is success. - Hardware: fact, assumption, benchmark, and route claim must be separate. - Auth: wrong-role denial and service-token behavior must be covered. - Routing: unknown friendly route rejects and records evidence.
Suggested cron names: uapi-hourly-diary-check, uapi-daily-rollup, uapi-promotion-verifier. Implementation may be cron, systemd timer, or Explorer-native automation, but must be visible through Explorer evidence.
Automation prompt starter: You are the UAPI continuation monitor. Use Explorer AI lane, verify readiness, check diary freshness and audit ids, publish a concise continuation packet if needed, and never perform hidden infrastructure work or print secrets.