Runbook Recall — resolve repeat Google Cloud tickets with sourced drafts

The problem

Platform teams repeat investigations, while past troubleshooting notes can be scattered or outdated relative to product guidance.

The idea

Why now

Google’s official skills catalog supplies product guidance, while claude-mem makes past agent-session observations searchable (google-skills-catalog, claude-mem).

What it combines

Google Agent Skills supply product-specific procedures; claude-mem supplies local investigation history. Source labels let engineers distinguish general guidance from a prior local fix.

How to build it

1. Choose a ticket type. Select a recurring issue and establish its historical resolution-time baseline. (1 day)
2. Approve sources. Select relevant Google skills, internal runbooks, and consented claude-mem exports; exclude sensitive material. (1 day)
3. Draft runbooks. Retrieve matching material and label proposed steps with sources and dates. (1 day)
4. Run a read-only pilot. Have engineers verify drafts and compare resolution times with historical tickets. (1 day)

Who benefits

On-call engineers

Spend less time repeating documented investigations.

Internal requesters

Get repeat cloud issues resolved sooner.

How it makes money

Model

Team subscription

Pricing

$499 per platform team per month

Wedge

One recurring Google Cloud ticket type at one platform team

Channels

Direct sales to platform-engineering directors · Rollout through the buyer’s on-call team

Distribution

A platform-engineering director pays. Start with one repeat ticket type and roll out through the existing on-call team.

Why it wins

Instead of an unsourced support answer, each proposed step identifies its guidance or internal-note source and date. Engineers verify every step; the assistant cannot remediate production systems.

Risks

Past agent sessions may contain little relevant incident history, and fixes may be stale. Bootstrap with approved runbooks, show dates, and test drafts against engineers’ existing process. Ingest only consented exports; do not intercept sessions or use unofficial platform access.

Strongest objection left standing

Teams without substantial relevant past agent sessions may gain too little from claude-mem to justify the subscription.

Build it with

• Google Agent SkillsSupplies official Google product skills for runbook drafting.
• claude-memProvides searchable observations from prior agent sessions when a team has them.

Repo to start from

runbook-recall - sourced drafts for repeat cloud tickets

Evidence

• Google Agent Skills
• claude-mem

Debated by openai/gpt-6-sol, anthropic/claude-sonnet-5.5, xiaomi/mimo-v2.6-pro, qwen/qwen3.8-max-0902, moonshotai/kimi-k2.5. Judged by JEV.