The patterns worth automating only become visible after a few weeks of doing the work. But by then, you’re too deep in the weeds to notice them.
The skill I built to solve this runs every Monday morning, reads the last seven days of my activity, and hands me a ranked list of the skills I should build next.
…yep. It’s a skill for skills. Whatever works, right?
A skill only pays off if you run it 5+ times a month. The build step is the easy part. Noticing what to build a skill around is harder (at least for me it was), and that’s where most operators stall. Memory is terrible at it. Your tools remember better than you do.
“If you can’t describe what you are doing as a process, you don’t know what you’re doing.”
W. Edwards Deming
👋 Hi, I’m Kaylee Edmondson. Looped In lands in your inbox every Sunday with one goal: to give you a sharper way to think about demand gen and growth in B2B SaaS. 2k+ marketers are already reading it. If you’re not subscribed yet, fix that below.
A One-Time Setup
Before the weekly audit ever runs, the skill onboards you with a 2-minute interview. Five questions, asked one at a time:
What’s your role, and which clients are you working on?
Which tools do you use most? (Slack, Gmail, Monday.com, Granola, Google Calendar, HubSpot, Google Drive, LinkedIn Ads, Google Ads, Salesforce)
What feels like it takes the most time in your week?
Do you already have any skills, templates, or automations set up?
What day of the week do you want the audit to run?
My answers got saved to a markdown file in my workspace called skill-audit-profile.md. Every future weekly run reads that file first. Self-reported time sinks get weighted heavier in the ranking. Existing automations get excluded from recommendations.
Then the skill schedules itself using a cron expression to fire at 9:00 AM on whatever day I picked. I picked Monday because I want the report waiting for me when I open my laptop at the top of a fresh week.
The Weekly Skill Audit: 3 Phases
Every Monday at 9:00 AM, the skill runs automatically. Three phases. I don’t have to prompt it. I don’t have to remember. By the time I’m pouring coffee, the report is sitting in my workspace.
Phase 1 is the scan. Phase 2 is the scoring. Phase 3 is the report.
Phase 1: Scan
The audit reads my profile first so it knows my role, my clients, my primary tools, and my self-reported time sinks. Then it scans the last 7 days of activity across whichever tools I marked as primary. Each tool has its own scan behavior:
Granola: queries recent meetings, scans titles and transcripts for patterns
Gmail: searches sent emails (
in:sent newer_than:7d) for templated language and recurring email typesSlack: searches recent messages for repeated phrases, status updates, recurring explanations
Monday.com: checks recent board activity and item creation patterns
Google Calendar: reviews the week’s events for recurring meeting types
HubSpot: checks recent CRM activity for repeated workflows
Google Drive: checks recently created or modified files for naming patterns
If a tool isn’t connected, the audit skips it and notes the gap in the final report.
Phase 2: Score
The audit looks for six specific pattern types. These are the categories baked into the skill:
Repeated creation (same type of deliverable made regularly)
Multi-step workflows (predictable sequences across tools)
Recurring communication (templated emails or messages)
Data gathering rituals (pulling info from multiple places on schedule)
Context switching overhead (bouncing between tools predictably)
Knowledge transfer moments (explaining the same thing to different people)
Every pattern that surfaces gets scored on five dimensions:
Frequency: how many times it showed up this week
Time per occurrence: estimated minutes per instance
Complexity: how complicated the workflow is
Automatable portion: what fraction a skill could realistically handle
Estimated weekly time saved: the ranking number
The patterns get ranked by estimated weekly time saved. The audit also cross-references every candidate against my profile. Self-reported pain points get weighted heavier. Anything I already have a skill for gets dropped from the list before it even hits the report.
The scoring works as a filter. It removes the candidates that would burn a weekend for low payoff. The patterns that survive the rank order are the ones worth a Monday of building work. Most weeks only one or two patterns make the cut. That’s correct.
Phase 3: Report
The audit writes a markdown file to my workspace called weekly-skill-audit-[YYYY-MM-DD].md. The structure is fixed:
# Weekly Skill Audit — [date]
**Tools scanned:** [list]
**Activity window:** Last 7 days
## New Skill Opportunities This Week
(Ranked by estimated weekly time saved. Only NEW patterns not flagged in
previous reports.)
### 1. [Skill name] — [est. weekly time saved]
**Pattern:** [what was observed]
**Evidence:** [paraphrased examples]
**What the skill would do:** [2-3 sentences]
**Complexity to build:** [Low / Medium / High]
## Recurring Patterns (Still Present)
(Flagged before, still showing up. Brief status update.)
## Gaps
(Tools not connected, insufficient data, light weeks.)
After saving the report, the skill sends me a short summary message in conversation. Three or four sentences, no more. Top one or two new findings only. If I want the details I open the file.
What the report does not include: a drafted SKILL.md file, trigger phrases, worked examples, or a ready-to-ship skill. Building the skills it surfaces is a completely separate job I do later in the week, with a different tool, only if I agree with the audit’s ranking.
I Ran It on Myself
A normal Monday. The week behind it: three ABM kickoff prep sessions, a stack of ad reviews, twelve Granola meetings, and one Slack thread where I’d spent half an hour walking a new team member through tier logic.
The audit covered Granola, Gmail, Slack, Monday.com, Google Calendar, and Google Drive.
The top opportunity from the report:
## New Skill Opportunities This Week
### 1. abm-strategy — est. 180 min/week saved
**Pattern:** Knowledge transfer moments + multi-step workflow + recurring communication
**Evidence:** 3 ABM kickoff prep sessions this week each working through
the same ~8 questions (tier structure, budget allocation, BDR trigger
thresholds, ad format mix by stage, intent signal handoff). One Slack
thread explaining the same budget math to a new team member. Two Granola
calls where the first 20 minutes were spent re-covering tier logic.
**What the skill would do:** Intake the 8 standard ABM questions, return
a draft tier structure, budget allocation per tier, BDR trigger rules,
ad format recommendations by stage, and an ads-to-outbound signal
framework. One shot, based on my existing ABM playbook.
**Complexity to build:** Medium
The summary message the skill sent me after the report saved:
“1 new opportunity worth ~3 hours/week: ABM strategy intake (Medium complexity, est. 180 min/week). Full report in workspace.”
The audit’s job stops at description. It captures the opportunity in 2-3 sentences, ranks for complexity, and shows me the evidence.
I built abm-strategy the following week. Took me about an hour, in part because the audit had already done the work of articulating what the skill needed to do.
The audit has since flagged four more adjacent patterns (quarterly ABM performance review, account list QA, tier migration logic, ABM-to-outbound feedback loop) that I’d never have connected to each other without seeing them stacked in one report. Each one earned its own line item in subsequent weekly audits.
What does this mean?
1. Discovery and building are different jobs.
The audit describes opportunities. It does not build skills. That separation is deliberate. When you mash discovery and building together, you prioritize whatever task you feel like automating in the moment, which is almost never the highest-leverage one. Forcing a beat between “this is worth building” and “I’m building it” gives you the chance to ask whether the audit’s #1 candidate is actually the one you should ship next. Sometimes I disagree with the rank. It can’t see into the future after all.
2. Your tools remember better than you do.
Your sent folder, your calendar, your call transcripts, and your Slack history are an honest record of your week. My memory is almost always a vibe check. The audit reads the actual artifacts, which is why it surfaces opportunities your brain would have missed.
3. The profile is the cheat code.
The one-time interview creates skill-audit-profile.md. Every future weekly run reads that file first. Self-reported time sinks get weighted heavier in the ranking. Existing automations get excluded from recommendations. Without the profile the audit still works, but with it, the audit picks skills you’ll build because they touch work you already admitted is painful. Two minutes of interview, six months of better recommendations. Worth it.
The objection I hear most often: “You need to be a developer to do this.” You don’t. SKILL.md is a markdown file with frontmatter. The audit’s scheduled task is set up for you by the skill itself. If you can answer five questions in an onboarding interview and pick a day of the week, you can run this. Writing SKILL.md is straightforward.
If you want this in your own setup, grab the /weekly-skill-audit SKILL.md template. Run the interview, pick your day, and let it scan your weeks.
And reply to this email with the skill you’re building next. I’ll feature the best replies in a future issue.
See ya next week,
Kaylee ✌
This is dope.
One question I couldn’t shake off was: How much in terms of tokens does this exercise take? Or how what plan are you on?
Just trying to gauge for all the scanning multiple transcripts, chats and documents….