· claude-code · consulting · skills · solo-consultants
Claude Code skills for consultants: the real list
An opinionated list of the Claude Code skills a solo consultant should actually run. What we ship, what's coming, and what you'll need to build yourself.
There are about a dozen things on a solo consultant’s plate that could plausibly be a Claude Code skill. Most aren’t. After building four of them into Knack’s first pack and discarding several more along the way, here is the real list: what’s worth turning into a skill, what’s coming next, and what you’ll need to build yourself.
The point of this post is not to sell you anything. It’s the curation we wished existed when we started. If you walk away and build all of this yourself, that’s a fine outcome.
What “Claude Code skills for consultants” actually means
A Claude Code skill is a markdown file that scopes the model to a single recurring job. It is not a chatbot, a workflow tool, or an agent in the autonomous sense. It is a prompt with discipline: when this fires, do this, ask for these inputs, refuse to invent these outputs, write to this format.
For a solo consultant, the skills worth writing share three traits. They run weekly or more (cost recovers). They produce a deliverable a client sees (quality matters more than speed). And they have a structure that’s stable across engagements but tedious to repeat (the win is in not re-deciding the structure each time). If a job is occasional, internal-only, or genuinely bespoke each time, a skill is the wrong tool.
That filter rules out about half the things people instinctively want to automate.
The four that hurt every week
These are the skills that make it into Knack’s Solo Consultant Ops pack. Each one fires on a job we’d been doing manually for years and were tired of re-deciding.
proposal-writer. Interviews you about a prospect, then produces a 3–5 page proposal in markdown. The structural opinions matter more than the prompt. Problem stated before credentials, one named risk, a single number rather than a range, a dated next-step CTA. The skill pushes back on vague answers and refuses to invent qualifications, deliverables, or timelines you didn’t give it. The hardest part of writing a good proposal is having the discipline to interview yourself; the skill enforces it.
sow-templater. Consumes the output of proposal-writer and produces the SoW: assumptions, dependencies, change-control mechanism, payment terms. SoWs are where most engagements quietly go wrong, and where most consultants under-protect themselves. A templated skill catches the omissions a tired Sunday-evening human will miss.
follow-up-sequencer. Generates a four-touch cadence after a discovery call, a proposal send, an end-of-engagement, or a quarter-end retention check. The cadence is the boring bit; tone-matching to the prospect’s communication style is the bit that makes them actually read it. Skills are good at the second once you give them the first as a fixed structure.
invoice-chaser. Drafts the T+0, T+7, T+14, T+30 escalation emails for an overdue invoice: polite first, then firm, then stating the contractual remedy. Most consultants under-chase because the emails are emotionally costly to write. A skill does the writing; you do the sending.
These four cover most of what bleeds time and money. They’re also the four with the cleanest input/output contracts, which is why they made the cut.
Two more coming in v1.1
The original plan was five skills plus an orchestrating agent. We cut it to four to ship sooner. Both deferred items are still on the roadmap.
rate-card-calculator. Given desired take-home, working weeks, utilisation rate, and overhead, computes a defensible day rate plus tiered packages: hourly, day rate, weekly retainer, fixed-scope. Most solo consultants set their day rate by feel, anchor low, and don’t revisit it. A calculator with the right inputs prevents that. It got cut from v1.0 because we wanted to ship four polished skills rather than five rough ones.
engagement-coordinator agent. Orchestrates the four skills into a single chat flow. I have a new prospect at $stage, what do I do next? The skills work fine standalone, but a thin coordinator on top removes the cognitive overhead of remembering which skill fits which moment. Deferred to v1.1 because agents add testing surface area, and the four standalone skills had to earn their keep first.
Three you’ll need to build yourself
These didn’t make the cut for the pack but matter enough that you should write them yourself.
Meeting-notes summariser. Transcript in (Otter, Granola, Read.ai), structured notes out: decisions, action items with owners, open questions. The reason this isn’t in the pack: every consultant uses a different transcription tool with different output formats, and the skill has to be rewritten for each. Write your own against your stack; it’s a 30-minute job.
Weekly status writer. Pulls the week’s deliverables, blockers, and next-week plan into a tight client email. Same problem as above. Every engagement has different status conventions, retainer terms, and reporting depth. The right shape is one skill per long-running client, not one generic skill.
Kickoff-doc generator. First-week onboarding artefact: RACI, comms cadence, document repository structure, escalation paths. High value but only fires once per engagement, so it earns less than the four above. Build it the third time you have to write one from scratch.
Where this approach breaks
A skill is a force multiplier on a job you already know how to do. It is not a substitute for the judgement that produced the structure in the first place. If your proposals already convert and your SoWs already protect you, a skill makes them faster. If they don’t, a skill writes worse versions faster, which is worse, not better.
Two specific failure modes worth naming. Skills hallucinate confidently. Proposals will invent client names, deliverables, and timelines if you don’t pin the inputs hard. Always read every word before sending. And skills compress; they don’t replace. The relationship-building work that wins consulting engagements still happens between humans on calls, not between a model and a markdown file.
The real list is shorter than most listicles suggest, because most consulting work is judgement, and judgement does not yet automate.
What this looks like with Solo Consultant Ops
The four skills above ship today in Solo Consultant Ops for a £99 lifetime licence: proposal-writer, sow-templater, follow-up-sequencer, invoice-chaser, plus templates and two worked examples. v1.1 adds the rate-card calculator and the orchestrating agent. The three skills you’ll build yourself are yours.
If the pack pays for itself in the first proposal you don’t have to write from scratch, that’s the bar.