Diagnostic analyst
What it does
Give it a website and your social accounts, and it writes a diagnosis of where your marketing stands and what to do in the next 30 days.
- Writes the diagnosis behind every check-up: brand overview, positioning, SWOT, channel recommendations, content strategy, competitive landscape, search intent and campaign ideas, and a strategic opinion. Search intent and campaign ideas are shown at the end of the channel recommendations.
- Prescribes a role, priority, content ideas, and KPIs for each channel, and recommends long-form formats.
- Lays out the 4-week plan (the cronogram) that seeds your first follow-up cycle.
- Reads your metrics summary for a date range and writes a short read-out: summary, insights, recommendations, and its confidence.
- Writes the diagnosis stage of the one-phrase Audit <domain> run, which produces the same analysis from a domain name alone.
What it reads
- The pages of your website, read live when the audit runs. The audit stops if the site cannot be read.
- The LinkedIn URL, Instagram handle, and goals you enter in the audit form.
- Public web research about your niche and competitors.
- Your business profile, and Google Analytics or Google Ads signals when those are connected.
- For the metrics read-out: the follower, impression, engagement, click, lead, and conversion figures PULSE has synced.
What it produces
A structured diagnosis, saved with the sources it used. Each section is stored from its own heading, so a section that mentions another section’s topic is kept whole. PULSE then copies its findings into the rest of the app: draft content, a draft campaign, a follow-up cycle, audience profiles, and a market summary. See Audit a domain.
The relevance audit beside it is not written by this agent. Its 0–100 scores come from a rule-based scoring engine.
A diagnosis runs to one end-to-end deadline — under three minutes by default — and every phase inside it is budgeted from what is left. A phase that does not finish is named on the card as not completed within that budget, and the run ends with a partial result. Text written before the budget ran out is not thrown away: it goes through every safety check the same way a finished draft does, and is shown as the partial answer it is, never as a finished analysis.
That deadline covers the diagnosis. The multi-stage one-phrase audit runs the diagnosis as one of its stages and bounds the run as a whole far more generously, detached from the request that started it.
Human oversight
- The diagnosis is advice. You read it and decide what to act on.
- The drafts it seeds wait in Drafts to approve on Campaigns, and the campaign it seeds is a draft. The 4-week plan queues scheduled posts only for channels an administrator has already armed, and a queued post goes out only after you approve its content. For a disarmed channel nothing is queued at all — the task stays in Your plan and its draft in Drafts to approve.
- Channel prescriptions and brand-visibility insights can be withheld by the plan gate when the judge disagrees with them. A withheld prescription never becomes a draft.
- It cannot publish, move budget, or change a campaign.
Model
The deployment’s base model, shared with every other PULSE agent — on Runink Server, a Qwen-family model under the Apache-2.0 licence. See Sovereign inference.
Where it runs and data handling
On the deployment’s own inference server: Runink’s shared machines on a Trial, your own cloud on Cloud and your own servers on On-premises. No outside AI service is used on any plan. Reading your website and researching your niche means PULSE fetches public web pages while the audit runs.
Guardrails
- Never invents market or competitor figures. PULSE refuses requests to fabricate them.
- No targeting or analysis by protected characteristics.
- It recommends; it never publishes or spends.
- Your website and web research are read as material, never as instructions.
- Follows the shared value-first brand rule and writes in the language you use.
Getting good results
- Give it every channel you want analysed: it works from your website and the profiles you list in the audit.
- Competitor observations come from competitors’ public pages.
- Its 0–10 channel scores are judgements to guide priorities, not measurements.