acrid-auto/skills/ops-manager

Ops Manager Skill

name: ops-manager
description: Keeps the Launch Cockpit, Skill & Agent Roadmap, and Scoreboard current. Runs at the end of every session alongside the Kaizen entry.

Purpose

Narrow operational specialist. One job: ensure Acrid's internal dashboard documents reflect reality at the end of every working session.

Not a content writer. Not a strategy tool. Not a Kaizen replacement. This skill updates the operational layer — the pages the operator looks at to know what's true right now.

---

Why This Exists

The Kaizen Log stayed current because it has a ritual (end of every session, Claude appends an entry). The Launch Cockpit and Roadmap went stale for 6 days because no process ensured they got updated. Same problem, same fix: make it a ritual with a skill that enforces it.

If the Cockpit "Updated" date is more than 2 days old, something is broken. This skill exists to prevent that.

---

When It Runs

End of every daily session, after the Kaizen Log entry is written and before the session closes. The Ops Manager runs last — it's the "save state" step.

Also runs on-demand when the operator asks to update the dashboard or roadmap.

Auto-trigger rule: If Claude notices the Launch Cockpit "Updated" date is more than 2 sessions old, flag it to the operator and offer to run the Ops Manager immediately.

---

Target Documents

This skill updates local operational files (repo-native, no Notion dependency):

1. Launch Cockpit — infrastructure/launch-cockpit.md
2. Skill & Agent Roadmap — infrastructure/roadmap.md
3. This Session's Kaizen Cross-Check — verify the Kaizen entry was written in memory/kaizen-log.md (do not duplicate it)

---

What Gets Updated

Launch Cockpit

| Section | Update Rule | | --- | --- | | "Updated" date | Set to today's date. Every session. No exceptions. | | Right Now — This Week | Move completed items to Done. Add new blockers/priorities that emerged today. Remove items that are no longer relevant. | | Done — Shipped & Working | Add anything that shipped or became operational today. | | Next Up — Priority Order | Reorder if priorities shifted. Add new items. Remove items that moved to Done. | | Scoreboard | Update any metric that changed (posts count, products, revenue, pipeline status, follower count if known). | | Decision Log | Append one line for today if a significant decision was made. Not every session needs a decision log entry — only when something meaningful was decided. | | Key Links | Add new links if new products/pages/repos were deployed. |

Skill & Agent Roadmap

| Section | Update Rule | | --- | --- | | "Updated" date | Set to today's date. | | Live Skills | Add newly completed skills. Update status notes if a skill's capabilities changed. | | Next Skills / Future Skills | Move items between tiers if priority shifted. Add new skill ideas that emerged. | | Automation Pipeline Status | Update blocker list. Move resolved blockers to "What's Working." Add new blockers. | | Product Status table | Update status, URLs, and prices for any product that changed today. | | Timeline | Check off completed items. Add new items for current week. Ensure next week's items are realistic. | | Tooling table | Update status column if any tool's status changed. |

---

How It Works (Step by Step)

1. Read both files — Read infrastructure/launch-cockpit.md and infrastructure/roadmap.md.
2. Diff against today — Compare what the files say vs. what actually happened this session. What shipped? What broke? What's new? What's obsolete?
3. Build the update list — For each file, list every change needed. Be specific: "Move 'permanent Cloudflare tunnel' from This Week to Done" not "update the cockpit."
4. Execute updates — Edit the markdown files directly using the Edit tool.
5. Verify — Re-read both files after updating to confirm changes landed correctly.
6. Report — Tell the operator what was updated in 2-3 sentences. No long explanations.

---

Rules

• Never invent progress. Only mark things as done that actually happened. If something is partially done, update the description to reflect partial status — don't move it to Done.
• Never delete decision history. The Decision Log is append-only. Old entries never get edited or removed.
• Preserve child pages/databases. The Launch Cockpit has embedded databases and child pages. When using replace_content, always include the child page/database references from the existing page.
• Keep it surgical. Prefer update_content (search-and-replace) over replace_content when only a few things changed. Replace_content is for major rewrites only.
• Scoreboard accuracy matters. If you don't know the exact number (e.g., follower count), write "TBD" or "check analytics" — never guess.
• The Kaizen Log is NOT this skill's job. This skill verifies the Kaizen entry was written. It does not write it. The Kaizen entry is part of the daily workflow, not the Ops Manager.
• Dates are not optional. Both pages must show today's date in their "Updated" line after this skill runs.

---

Failure Conditions

The Ops Manager has failed if:

• The "Updated" date on either page is more than 2 days old
• The Scoreboard shows metrics that are demonstrably wrong
• A shipped product/feature is still listed as "not started" or "next"
• A known blocker isn't listed in the Roadmap's blocker section
• The Timeline has items checked that aren't actually done
• Child pages or databases were accidentally deleted during a replace_content
• The operator opens the Cockpit and it doesn't match reality

---

Pre-Run Checklist

Before running, confirm:

1. ☐ Kaizen Log entry for today is written (or about to be written)
2. ☐ You know what was built, shipped, broken, or decided this session
3. ☐ You have the current state of both target files (read them)
4. ☐ You have a specific list of changes to make (don't wing it)
5. ☐ Drift check: Did any workflow, product, tool, or infrastructure change this session? If yes, verify that ALL referencing skill files and CLAUDE.md were updated. Grep for stale references (old tool names, old URLs, old product counts, "Notion" in posting paths). Flag anything missed.

---

Integration with Daily Workflow

The daily session flow:

1. Open chat → Claude reads Kaizen + databases + skills
2. Produce content (threads, posts)
3. Work on day's build
4. Write DITL blog post
5. Write Kaizen Log entry
6. Run Ops Manager ← this skill
7. Session closes

Steps 5 and 6 happen back-to-back. The Kaizen captures lessons. The Ops Manager captures state. Both are non-negotiable end-of-session rituals.

---

Built to prevent drift. Updated every session. The only skill whose success metric is that you never notice it.

CHECKLIST.md

LEARNINGS.md