The Pace cookbook

Install the sales plugin

Pick one path; both end up with the same plugin installed.

Authorize HubSpot and Slack

The first time you run any sales skill, Claude opens an OAuth window. Sign into HubSpot, then Slack. Once per service.

Create the scheduled routine

Go to claude.ai/code/routines → New routine. Pick Schedule trigger, set it to Weekly · Monday · 06:30 in your timezone.

Paste this as the routine prompt

This is the actual prompt Claude runs each week. Edit the channel name and account count to match your team.

Run /sales:pipeline-review against HubSpot. Slice the
output by deal stage. For each of the top 10 accounts by
deal size, spawn a sub-agent that runs /sales:account-research
with that account's name. Roll up all sub-agent findings into
a single Monday morning brief: pipeline summary at the top,
per-account notes underneath. Post the brief to #sales-pipeline
in Slack.

Click Run now to verify

The first run takes 2-3 minutes. The brief lands in #sales-pipeline. Once you've seen it work, the routine fires every Monday at 06:30 automatically. No further action needed.

Install marketing + data plugins

Marketing has the report skill; data has the SQL skill it leans on.

Authorize Amplitude, HubSpot, and Slack

The first time you run any marketing or data skill, OAuth windows open one by one. Sign into each once. Amplitude is the analytics source; HubSpot is the CRM; Slack is where the digest lands.

Create the scheduled routine

Go to claude.ai/code/routines → New routine. Pick Schedule, set to Weekly · Friday · 16:00.

Paste this as the routine prompt

Edit the channel name and the metrics list to match your team's KPIs.

Run /marketing:performance-report covering the past 7 days.
Pull funnel metrics from Amplitude (signups, activations,
free-to-paid conversion) and lead-source attribution from
HubSpot. Use /data:write-query if you need to grab anything
the report skill doesn't have directly. Format as a Slack
digest: headline numbers at the top, week-over-week deltas,
top 3 campaigns by impact, anything anomalous worth a closer
look. Post to #marketing-weekly.

Click Run now to verify

First run takes a few minutes (it has to authenticate Amplitude on the cloud side too). The digest lands in #marketing-weekly. Future Fridays at 16:00 are automatic.

Install the engineering plugin

Brings /engineering:code-review and 33 other skills.

Install the Claude GitHub App on your repo

GitHub event triggers need the Claude GitHub App. The routine setup at claude.ai/code/routines prompts you to install it on the relevant repo if you haven't yet.

Create the GitHub-triggered routine

At claude.ai/code/routines → New routine. Pick GitHub event, choose your repo, event pull_request.opened. Optional filter: is draft = false so it only fires on ready-for-review PRs.

Paste this as the routine prompt

A new PR was just opened in this repo. Run /engineering:code-review
on the PR's diff. Spawn one sub-agent per changed file for parallel
review. Focus on: bugs, naming, dead code, weak tests, convention
drift from neighboring files. Post findings as DRAFT review comments
on the PR (not approvals) so the author can accept or dismiss each
one before tagging a human reviewer. Skip whitespace nits.

Open a test PR to verify

Make a tiny change on a branch and open a PR. Within a minute or two the routine fires and draft review comments appear on the PR. From here on, every new PR gets the same treatment.

Install the data plugin

Authorize your warehouse + Slack

Snowflake / Databricks / BigQuery use Custom URL authorization : provide your account's MCP endpoint the first time Claude asks. Slack is OAuth.

Create the scheduled routine

At claude.ai/code/routines → New routine → Schedule · Weekly · Monday · 07:00.

Paste this as the routine prompt

Edit the KPI list to match your team's metrics + the channel name.

Run /data:statistical-analysis against last week's data
for these KPIs: signups, activations, weekly active users,
revenue, gross retention. Use /data:write-query as needed
to pull the underlying numbers from the warehouse. Flag
any metric that moved more than ~2 standard deviations
from the trailing 8-week baseline. For each anomaly,
include: the slice (segment / cohort), the SQL that
surfaced it, week-over-week delta, and one candidate
explanation. Post the digest to #data-weekly in Slack.

Click Run now to verify

The first run may take 5+ minutes if your warehouse queries are heavy. Verify the digest format, then trust the Monday 07:00 cadence going forward.

Install the product-management plugin

Authorize Fireflies or Otter

Whichever transcript service you use. First time you ask Claude to read a transcript, OAuth opens. Optional: Notion or Linear if you want the synthesis doc saved there.

Open a Claude session with your interview list

No routine needed : this one is interactive (you only run it at the end of each sprint). In Claude Code or Cowork, start a new chat.

Paste this prompt with your sprint's transcript list

Run /product-management:synthesize-research against these
interviews from the past sprint:

  - <Fireflies URL or meeting title 1>
  - <Fireflies URL or meeting title 2>
  - <...>

Spawn one sub-agent per transcript to extract themes,
direct-quote evidence, and friction points. A coordinator
sub-agent should roll them up into a single synthesis doc:
top 5 themes by frequency, supporting quotes per theme,
disagreements between participants, and 3 candidate
product implications. Use plan mode : show me the
synthesis before saving.

Review and save

Plan mode parks the draft for your review. Edit themes, push back where the AI over-synthesized, then save to wherever your team keeps research docs (Notion, Linear, Confluence).

Install the finance plugin

Authorize your warehouse + Microsoft 365

Snowflake / Databricks / BigQuery for the ledger pull; Microsoft 365 for Excel writeback. Custom URL prompts on first run.

Create the monthly routine

At claude.ai/code/routines, preset schedules don't include "last business day of the month." Create the routine with the closest preset, then run /schedule update in the CLI to set a cron expression like 0 8 28-31 * 1-5 (8am, last few weekdays of the month).

Paste this as the routine prompt

Run the month-end close in order:

  1. /finance:close-management to orchestrate the checklist.
  2. /finance:reconciliation against the warehouse ledger
     (call out any unreconciled balance).
  3. /finance:variance-analysis vs plan + prior month.
  4. /finance:journal-entry-prep for any accruals or
     reclasses the variance analysis surfaces.

Use plan mode for the journal entries : do NOT post
anything to the GL automatically. The controller will
review and approve each entry. Post a close-progress
summary to #finance-close.

Manually run for this month's close to verify

Click Run now at the end of this month. Walk through the journal entries in plan mode; approve, edit, or reject each. Once you trust the output, the routine fires on the last business day automatically.

Install the legal plugin + write your playbook

The plugin ships with skills; the playbook is your team's standard positions on common clauses (e.g. mutual NDA, IP assignment, liability caps).

Set up the playbook in a docs folder

Run /engineering:setup-engineering-skills once at your repo or workspace root. It scaffolds docs/agents/legal-playbook.md for you. Drop your team's standard positions on common clauses there. The legal skills read from this on every call.

Create an API-triggered routine

At claude.ai/code/routines → New routine → API trigger. After saving, click Generate token and copy both the URL and the bearer token (shown once).

Wire your inbox to POST inbound NDAs

Set up a Gmail/Outlook forwarding rule (or Zapier) that POSTs new emails with subject containing "NDA" to your routine's /fire endpoint, with the email body in the text field of the JSON payload. Paste this as the routine prompt:

An inbound NDA just arrived. The email body is in the
incoming text field. Run /legal:triage-nda to classify
it (mutual / one-way, standard / non-standard, urgency).

If the triage says "standard": run /legal:review-contract
with our playbook in the context. Produce a redlined
version as a draft response with explanatory comments.

If "non-standard": post a Slack message in #legal-queue
flagging the specific clause(s) that need a human and
the urgency level. Do NOT auto-redline non-standard
agreements.

Test with a real (sample) NDA

Forward a sample NDA to the configured address. Within a couple minutes you'll either see a redlined draft (standard) or a flag in #legal-queue (non-standard). Once you trust the triage classifier, the queue stays drained automatically.

Install the operations plugin

Authorize Linear (or Jira), Slack, Notion

Whichever ones your teams actually use. OAuth opens for each on first use.

Create the scheduled routine

At claude.ai/code/routines → New routine → Schedule · Weekly · Friday · 14:00.

Paste this as the routine prompt

Edit the team list to match your org structure.

Run /operations:status-report. Spawn one sub-agent per
team to pull progress signals from the last 7 days:

  - Engineering: closed Linear tickets, merged PRs, incidents
  - Product: spec updates, sprint progress
  - Sales: closed-won, pipeline movement (top-line)
  - Marketing: shipped campaigns, performance
  - Customer Support: ticket volume, escalations

A coordinator sub-agent rolls them into one Friday digest:
"what shipped" at the top, blockers + risks in the middle,
next-week focus at the bottom. Post to #leadership-friday
in Slack and mirror to docs.pace/status/<YYYY-MM-DD>
in Notion.

Click Run now to verify

Read the first digest carefully : adjust the per-team scope until the rollup matches how your org actually thinks about "progress". Once it's right, Fridays at 14:00 are hands-off.

Install the human-resources plugin

Authorize your ATS + Slack

Whichever ATS your team uses (Ashby, Greenhouse, Lever, etc.). If your ATS doesn't have a first-party MCP server yet, see the connector docs for the Custom URL pattern. Slack is OAuth.

Create the scheduled routine

At claude.ai/code/routines → New routine → Schedule · Weekly · Monday · 08:00.

Paste this as the routine prompt

Run /human-resources:recruiting-pipeline against the
ATS. For each open requisition, surface:

  - Stage of every active candidate
  - Time-in-stage (flag anyone > 10 business days)
  - Interviews scheduled this week (panel + candidate names)
  - Offers extended / accepted / declined since last week
  - Reqs without any candidates in pipeline ("stuck")

Format as a digest grouped by hiring manager. Post to
#hiring-weekly in Slack. Tag each hiring manager next
to their reqs so they get a heads-up DM.

Click Run now to verify

Eyeball the first digest for accuracy (does aging match what you see in the ATS?). Once you trust it, Mondays at 08:00 are automatic.

Install the customer-support plugin

Authorize Intercom (or Zendesk), HubSpot, Slack

Intercom / Zendesk for the ticket queue; HubSpot for account context; Slack for escalations.

Create an API-triggered routine

At claude.ai/code/routines → New routine → API trigger. Copy the URL + bearer token after save.

Wire Intercom (or Zendesk) to POST new conversations

In Intercom: Settings → Webhooks → New webhook, subscribe to "conversation.user.created", point at the routine URL with the bearer token. (Zendesk: Apps & Integrations → Webhooks.) Then paste this routine prompt:

A new support ticket arrived. The conversation body is
in the incoming text field.

  1. Run /customer-support:ticket-triage to classify
     urgency (P1/P2/P3) and category.
  2. Pull the customer's account context from HubSpot
     (plan tier, recent activity, prior tickets).
  3. Run /customer-support:draft-response : tone
     should match the segment (enterprise / SMB / free).
     Park the draft in the conversation as a private note
     for the human agent to review before sending.
  4. If P1 OR enterprise tier: run
     /customer-support:customer-escalation to ping the
     on-call engineer in #support-escalations.

Send a test ticket to verify

Open a test conversation in Intercom from a sandbox account. Within ~30 seconds you should see a private note with the draft response. Once the format is right, every new ticket gets the same treatment.

Install the productivity plugin

Authorize Slack, Linear, Notion, and GitHub

Whatever subset you use. The skill pulls signals from each, so the more sources connected the richer the wrap.

Create the scheduled routine

At claude.ai/code/routines → New routine → Schedule · Weekdays · 17:30.

Paste this as the routine prompt

Edit the channel + role-focus line to match your team.

End of day check-in. Run /productivity:update to assemble
what I shipped today: commits merged, Linear tickets I
moved, Slack threads I contributed substantively to, docs
I edited in Notion.

Then run /productivity:task-management to roll my open
items into tomorrow's queue (carry forward unchecked
todos, surface things I said I'd do but haven't).

Draft a 3-line async standup post in our voice:
"Yesterday I... Today I will... Blockers: ..."

Use plan mode : show me the draft for a quick edit
before posting to #async-standup.

Click Run now to verify

The first run today shows you the draft format. Plan mode means you approve before it posts. Tomorrow at 17:30 it runs again automatically : you'll get a notification in the Claude app when the draft is ready for your edit.