Automation Register — Row Template

Module 6 capstone scaffold · introduced in Lesson 6.1 · concept shape + recipe path fields

What this is. The header block and row template for automation-register-v1-draft.md, the capstone artifact Module 6 grows across all five lessons. Copy the header block once. Copy the row template once per task, in the order the lessons build them.

How to use. Paste the header block into a new file called automation-register-v1-draft.md at the start of Lesson 6.1. At each lesson’s project checkpoint, add one row. Fill every column before the task is allowed to run. By Lesson 6.5’s freeze, you will have five rows plus a retired-entry row for the retirement ritual.

The load-bearing rule of Module 6 — keep this visible while you fill in rows.

Audience = only you. No row is allowed in this register with a different value.

Header block — copy once into automation-register-v1-draft.md

# Automation Register — v1 (draft) Owner: <your name> Started: <YYYY-MM-DD> Home time zone: <America/New_York, etc.> ## Posture reference Audience for every task in this register: me only. Per-task cost ceiling: <$X/run, $Y/month>. Total-suite monthly ceiling: <$Z>. Maximum next-review interval: 90 days.

Row template — use once per task

## Entry <N>: <Task name> - Task ID: <slug; matches folder name under ~/ai-architect-academy/automation/> - Purpose: <one sentence — what question this task answers for me> - Schedule: <trigger, e.g., “Daily 06:30 America/New_York”> - Scope / credentials: <the narrowest inputs and permissions the task needs; name each OAuth scope, API key, MCP connector> - Idempotency key: <what makes a re-run produce one artifact, not two — date, ISO week, query ID, event ID, content hash> - Artifact path: <exact file or folder, e.g., ~/ai-architect-academy/automation/briefs/brief-YYYY-MM-DD.md> - Audience: me only ← required value in Module 6; if not, this row is not allowed - Per-run cost (estimate): $<N.NN> - Per-run cost (actual, most recent): $<N.NN> - Monthly cost (rate): $<N.NN/month> — computed as per-run × runs-per-month - Cost ceiling: $<N.NN/run>, $<N.NN/month> - Last run: <YYYY-MM-DD HH:MM> - Last success: <YYYY-MM-DD HH:MM> ← successful AND student read/used the artifact - Last fire: <YYYY-MM-DD HH:MM> ← watchers only; blank for briefs/reports/refresh - Next review: <YYYY-MM-DD> ← within 90 days - Retirement trigger: <the specific real-world event that would make this task obsolete> - Prompt-change log: - <YYYY-MM-DD — one-sentence reason for a change> - <YYYY-MM-DD — one-sentence reason for a change> - Status: Active | Redesigning | Paused (past review) | Retired YYYY-MM-DD ### Notes <free-form notes — what the task is learning to do well, what it still gets wrong, anything future-you will want>

Rules the register enforces

  1. Every row has every column filled in. A column labeled “—” is acceptable for fields that do not apply (Last fire for non-watchers); a blank or missing column is not.
  2. Audience = me only. Any row with a different value is not allowed in Module 6.
  3. Idempotency key is a single specific token that names the artifact. “Date” is not a key; brief-YYYY-MM-DD.md using TZ=America/New_York date +%Y-%m-%d is.
  4. Next review is within 90 days. A past-due next-review date pauses the task until the student audits it (see the retirement ritual worksheet).
  5. Cost ceiling is both per-run and per-month. Both matter: per-run catches a single expensive pathological run; per-month catches slow drift.
  6. Prompt-change log is append-only. One line per change. Do not silently edit prompts without a log entry, and do not delete prior log entries.
  7. A Retired row stays in the register with a post-mortem. Deletion loses history; the register is the institutional memory of your automation suite.

Example filled row (reference only; use your own values)

## Entry 1: Morning brief - Task ID: morning-brief - Purpose: One-page summary of yesterday’s agent-access inbox and today’s Agent Access calendar, read at breakfast. - Schedule: Daily 06:30 America/New_York - Scope / credentials: Gmail label `agent-access` (read-only OAuth), Google Calendar `Agent Access` (read-only for conflict-check on `primary`). - Idempotency key: date in America/New_York — filename `brief-YYYY-MM-DD.md` - Artifact path: ~/ai-architect-academy/automation/briefs/brief-YYYY-MM-DD.md - Audience: me only - Per-run cost (estimate): $0.10 - Per-run cost (actual, most recent): $0.09 - Monthly cost (rate): $2.70/month - Cost ceiling: $0.25/run, $10/month - Last run: 2026-04-22 06:30 - Last success: 2026-04-22 06:31 - Last fire: — - Next review: 2026-06-21 - Retirement trigger: I have read fewer than 2 of the last 10 briefs. - Prompt-change log: - 2026-04-20 — tightened section 2 to require at least one non-newsletter item. - 2026-04-21 — re-stated honesty rails after section 5 produced “no issues” two days running. - Status: Active ### Notes Section 2 is the part I actually read. Section 3 has not caught anything the morning scan wouldn’t; consider scoping narrower next review.

Print this page. Paste the header block into your draft register. Add one row per task, in the order the lessons build them.