One Hundred Workflows, Sorted
A viral grid promises "100 agentic Claude workflows — automate your entire business operations." We audited the hundred like a directory instead of admiring the number. Honest inventory: about ten families of work, heavy duplication, a minority that is genuinely agentic — and inside all of it, a dozen items truly worth building. Here they are, each mapped to the mechanism that actually runs it.
01 — The arithmeticWhat a hundred items compress to
Line the hundred labels up and families emerge immediately. "Meeting Note Summary" (#27) and "Meeting Minutes Extraction" (#70) are one job. "Competitor Ad Scraper" (#6), "Competitor Pricing Watch" (#33), "Competitor Feature Gap" (#84) and "Brand Health Monitor" (#90) are one competitive-intelligence practice wearing four names. Sorted by the job rather than the label, the hundred collapse to roughly ten families:
| Family | What it covers | ≈ items |
|---|---|---|
| Content production | posts, scripts, hooks, captions, repurposing | ~16 |
| SEO & discovery | keywords, briefs, landing copy | ~6 |
| Competitive & market intel | scrapers, watches, gap analyses, trend tracking | ~9 |
| CRM & pipeline hygiene | lead scoring, data cleaning, pipeline audits | ~8 |
| Outreach & follow-up | sequences, personalization, reminders | ~8 |
| Code quality & CI | refactoring, tests, reviews, pipeline scripting | ~12 |
| Docs & knowledge | documentation, wikis, SOPs, knowledge search | ~8 |
| Inbox, calendar & admin | sorting, scheduling, expenses, invoices | ~10 |
| Finance & reporting | budgets, models, reports, dashboards | ~8 |
| Strategy artifacts | SWOTs, roadmaps, pitch decks, projections | ~15 |
None of this makes the poster useless — as a brainstorming menu it's decent. It makes the number a marketing device. Directories are honest when they count jobs, not labels.
02 — The word "agentic"Most of these are prompts
A workflow is agentic when it has a trigger (it starts without you), tool access (it reads or writes real systems), procedure (more than one step), and verification (it checks its work). By that test, most of the hundred are one-shot prompts — "SWOT Analysis Engine" is a well-dressed question. The genuinely agentic minority: inbox sorting, scheduled monitors and scrapers, CRM data cleaning, CI pipeline work, knowledge-base upkeep — the items that need MCP connections, schedules, or hooks to exist at all. Both kinds are useful. Only one kind is automation.
03 — The twelve worth building firstEach mapped to its real mechanism
- Inbox triage — scheduled task + email MCP; flag what needs you, draft the routine replies.
- Meeting notes → CRM — one skill: transcript in, summary + action items + CRM update out.
- Competitive watch — scheduled task + web access; weekly diff of competitor pages, not a daily scrape you'll never read.
- Content repurposing — skill anchored to your top posts (the working version is sheet C-12).
- Lead scoring & CRM hygiene — MCP to the CRM + a rules skill; run weekly, log what it changed.
- Follow-up sequencing — scheduled task reading the pipeline, drafting (never auto-sending) the nudges.
- Code review pass — subagent on a hook or CI step; findings as comments, human merges.
- Test generation — skill invoked per changed file; the suite is the verification.
- Documentation refresh — scheduled task diffing docs against merged changes; proposes, doesn't overwrite.
- Expense & invoice categorization — MCP to the books + a rules file; month-end run with an exceptions list.
- Report generation — skill + data MCP; same template every week so drift is visible.
- Knowledge-base search & upkeep — MCP over the wiki; answers cite pages, staleness gets flagged.
Every item above is trigger + tools + procedure + verification. That's the test the other eighty-eight mostly fail — and the test to apply to the next viral hundred.
04 — The ledgerThe poster's own claims, checked 2026-08-05
| Claim | Verdict | The record |
|---|---|---|
| "100 workflows" | NUANCE | ~10 families; heavy label duplication. |
| "Agentic" | NUANCE | A minority pass the trigger/tools/procedure/verification test; most are prompts. |
| "Automate your entire business operations" | NO SOURCE | Overreach; the twelve above automate real slices, with humans at the gates. |
| The underlying jobs are real Claude use cases | VERIFIED | Every family maps to shipping mechanisms — skills, MCP, scheduled tasks, hooks, subagents. |
| "Claude 4.6" as the current model | NUANCE | Dated — the Claude 5 family is current. Advice about products decays at product speed (see I-10). |
| "Comment for the master guide" | NO SOURCE | Engagement mechanic; the real guide is the documentation, public and free. |