Skills That Implement: the Cogny Solo Audit, Implement, Routine Loop
Skills That Implement: the Cogny Solo Audit, Implement, Routine Loop
Marketing audits have a shelf life problem. You run one, you get a solid list of findings, and then the list goes to die in a tab. The title tags never get rewritten. The burned-out ad keeps spending. The broken conversion event stays broken for a month and quietly poisons every report downstream.
The problem was never the analysis. It's the gap between knowing and doing.
So we rebuilt Cogny Solo's skills around closing that gap. A skill no longer ends with a report. It ends inside your accounts, with the fixes applied.
TL;DR
- 19 skills across 6 channels: 9 audits and 10 routines covering Google Ads, Google Search Console, GA4, Meta Ads, TikTok Ads, and Mailchimp. Every channel has an audit to start with and a routine to keep it tuned.
- Audits implement, not just recommend. The SEO audit stages title and content fixes in your CMS as drafts. The Google Ads audit pauses wasted spend and adds negative keywords. The GA4 monitor repairs broken tags in Google Tag Manager. The Meta and TikTok audits pause fatigued creative and move budget toward winners.
- Every write is approval-gated. Cogny asks before each change, stages GTM work as draft workspace versions, creates email campaigns as drafts it never sends, and moves budgets in steps. Nothing destructive happens without your explicit go-ahead.
- The hub keeps you in the loop. The Skills menu ranks skills by what users actually run, groups them per channel, shows "ran 6d ago" on each, and nudges you when a weekly routine is due. Locked channels connect right from the menu.
The loop
The design is a cycle, not a one-off:
- Audit. Connect a source and run its audit skill. You get a health score (0-100) and a ranked list of issues, every finding cited from your live data. No invented metrics, no generic advice.
- Implement. The audit ends by offering to fix the clear wins. You approve each change, Cogny executes it through the same connected account.
- Routine. A week later, the channel's routine skill picks up where the audit left off: catch the next fatiguing ad, the next striking-distance keyword, the next tracking break, before they cost you a month.
- Come back. The Skills menu tracks when you last ran each skill and puts a nudge on the button when a routine is due. If you drift away entirely, an optional email pulls you back to it.
What "implement" actually means, per channel
This is the part that matters, so here it is concretely:
| Channel | The audit finds | Cogny implements |
|---|---|---|
| SEO (Search Console) | low-CTR titles, striking-distance queries, decaying pages | stages the rewritten titles and content fixes in your CMS as drafts, resubmits sitemaps |
| Google Ads | wasted spend, irrelevant search terms, budget-capped winners | pauses the waste, adds the negative keywords, fixes the budgets |
| GA4 | conversion events at zero, unassigned traffic spikes | creates or repairs the tags in Google Tag Manager, staged as a draft version you review and publish |
| Meta Ads | creative fatigue, audience overlap, misallocated budget | pauses burned-out ads, adds converter exclusions, moves budget toward winners in steps |
| TikTok Ads | falling hook rates, rising CPMs, learning-phase traps | pauses fatigued creative, reallocates ad group budgets |
| Mailchimp | dormant subscribers, blast-only sending, deliverability risk | builds the engaged and dormant segments, tags the re-engagement cohort, drafts the weekly campaign |
The safety rails
Handing an AI write access to your ad account is only sane with hard rules, so every skill carries them:
- Go-ahead per write. Each change is proposed with the real figures behind it ("Frequency's at 3.1, CTR down 34%, want me to pause it?") and executed only on your yes.
- Drafts wherever drafts exist. GTM changes land in a draft workspace version, never auto-published. Email campaigns are created as drafts, never sent or scheduled. CMS edits stage as drafts.
- Steps, not swings. Budget moves happen in 20-30% increments. No skill ever pauses a whole campaign on a single signal.
- Irreversible means confirm-first. Archiving a dead email address or deleting anything requires an explicit per-item confirmation, and most skills simply refuse the destructive path.
The hub
All of it lives in the Skills menu, next to Recent chats in the report builder:
- Popular sits on top, ranked by what Cogny users actually run most (real run counts, not our guess).
- Below that, skills group per channel, audits first, routines after, each labeled with when you last ran it.
- Channels you haven't connected show as Connect to unlock: click one and the OAuth popup opens right there, and the channel's skills unlock in place when you're done.
- Once you've run an audit, a dot appears on the button when a routine is due, with a one-click "time to run" shortcut at the top of the menu.
The Get started panel beside the chat covers the same ground for first-time setup: connect a source, see what it unlocks, run your first audit.
What it costs
Skills are part of Cogny Solo at $9/month. Running them in the built-in chat uses your credit balance, and driving them from your own agent over MCP is unmetered under fair use on Solo, because your agent pays for its own inference. The free tier includes 15 MCP calls and a 10-credit chat gift, which is enough to connect a source and run your first audit before deciding anything.
If you'd rather script the whole loop from a coding agent, every skill's underlying tools are also callable through the Cogny CLI.
Start with the audit
Connect your biggest channel, run its audit, and approve the first fix. That single loop, finding a real problem and shipping the real correction in the same sitting, is the whole product thesis: analysis is cheap now. Implementation is the moat.
Get started free, or if your data is already at scale, take the 2-minute self-guided Cloud demo.