fable-agent/docs/receipts/factory-live-drift-20260621.md

Factory Live-vs-Memory Drift Receipt — 2026-06-21

Decision

DEPLOY DECISION: FAIL CLOSED
TRUST ORDER: live receipt > committed receipt > memory summary

The provided memory is useful historical context, but it conflicts with current live checks. For fable-agent deploy decisions, live evidence wins.

Memory claims provided

Claim	Memory value
Containers	25 Docker containers all UP
Agents	6 agents
Infrastructure	14 infra containers
Monitoring/deploy	5 monitoring/deploy containers
Deploy route	deploy-webhook on 8098; git-proxy serves /deploy
Skills	63 skills
RSI cron	skill_health.py runs every 6 hours
Auto-patch	rsi_canary degraded → repaired → recovered proof exists in pi-real-engineering

Live checks observed in this repo session

Commands run:

node dist/index.js factory check
node dist/index.js factory rsi --out .fable/rsi/quality-review-rsi.json
node dist/index.js fable5 receipt-health --root .fable --window-minutes 10080

Observed live facts:

Fact	Live value	Decision impact
Factory status	stale	fail closed for deploy readiness
Containers	14	conflicts with memory's 25
Agents	6	matches memory
Infrastructure	14	matches memory
Monitoring	5	matches memory
Skills	63	matches memory
Deploy route	8099/deploy present	route exists
RSI score	66/66 (100%)	healthy current score
RSI cron	absent / grep failed	conflicts with memory's cron-present claim
Factory watcher	active	positive
Auto-patch proof	not proven by current live RSI receipt	cannot claim generic autopatch
Receipt health	healthy	receipt consumer/homeostasis works

Reconciliation

Memory claim 25 containers all UP is stale against live 14 containers.

Memory claim skill_health.py runs every 6 hours is stale or at least currently unverified because live cron check failed.

Memory claim rsi_canary auto-patch proof exists is historical and should remain scoped:

Proven: rsi_canary degraded → patched → recovered in external receipt.
Not proven: generic non-canary degraded repair in current fable-agent receipt set.

Safe claims allowed now

• fable-agent code, tests, and docs are green.
• Memory B cell cache is implemented and tested.
• Receipt-health/homeostasis check is implemented and live healthy.
• 8099/deploy via git-proxy route exists.
• RSI current health is 66/66 (100%).
• Factory watcher is active.

Claims blocked until fresh receipts exist

• Factory fully fresh/green.
• 25 containers all currently up.
• Cron currently installed and active.
• Generic autonomous self-healing.
• Non-canary degraded repair.
• Deploy readiness when factory status reports stale.

Quality rule

If memory and live checks disagree, fable-agent must preserve both in the receipt and gate on live evidence.