R1 The engagement

From first call to
handbook in hand,
in seven steps.

How a typical engagement runs, what Claude does at each step, what we need from you, and what you walk away with. The same frame scales from a two week pentest to a six month product build.

R2 Visually

Every engagement runs on one of three cadences.

We only have three engagement shapes: a fortnight, a six-weeker, and an ongoing retainer. Everything else is a variant of these.

FORTNIGHT · 14 calendar days · for targeted scopes

Day 1 - 2Discovery
Day 3 - 7Research & work
Day 8 - 11Drafting
Day 12 - 13Review
Day 14Handover

SIX WEEKER · 42 calendar days · the default shape

Week 1Discovery
Week 2Research
Week 3Drafting
Week 4Drafting
Week 5Review
Week 6Handover

RETAINER · monthly cadence · for continuous work

Week 1Plan + kickoff
Week 2Execute
Week 3Execute
Week 4Ship + review
RollingHandbook update
R3 The seven steps, in detail

How it actually unfolds.

Step 01 · Fit call · 30 min

We decide together whether this is a fit.

You describe the problem. We describe how we work. If it maps to something we're good at and your timeline is realistic, we propose a scoping sprint. If not, we say so and point you at who is better suited. No sales follow-up.

What Claude does
  • Pre-call research: reviews your public site, funding, recent news
  • Generates a pre-call brief so the human operator arrives informed
  • Drafts the follow-up email with a summary of what we discussed
Signed NDA Fit memo Next step proposal
Step 02 · Scoping sprint · 3 to 5 business days

We scope the engagement with precision so you can say yes or no.

A short paid sprint (between $3k and $8k, credited against the engagement fee) where a senior operator plus Claude produce a written Scope Document: problem statement, success criteria, boundaries, risks, timeline, and price. You get a fixed-fee engagement letter at the end, not a range or a proposal with TBDs.

What Claude does
  • Ingests your existing artifacts (docs, code, decks) and generates a discovery summary
  • Drafts 3 possible engagement shapes with cost, timeline, and deliverable tradeoffs
  • Writes a first draft of the Scope Document for the operator to review and sharpen
Scope Document Three engagement shapes Fixed engagement letter
Step 03 · Kickoff · 2 to 3 days

We open the engagement, align the team, and set up the instrumentation.

Kickoff workshop (half day, remote or on-site). You introduce stakeholders, we share the working repository, the handbook draft location, the cadence of updates, the Slack (or whatever you prefer) channel, the escalation path, and the review calendar. We set up weekly 30-minute syncs and a daily async thread so you always know where we are.

What Claude does
  • Generates the working repository structure and onboarding doc for your team
  • Drafts the first version of the handbook table of contents
  • Sets up the engagement tracker with deliverables, owners, and dates
Shared working repo Engagement tracker Communication protocol
Step 04 · Research · 1 to 2 weeks

We do the deep work: research, recon, architecture review, or audit, depending on the service.

The shape of research depends on the service line. For security: authorized testing, cloud recon, threat modeling. For web and product: codebase review, architecture mapping, user and stakeholder interviews. For SEO/GEO and marketing: keyword and share-of-answer audits, competitor benchmarking, ICP research. Every finding is captured in the handbook draft in real time, not written up later.

What Claude does
  • Runs automated discovery against agreed scope (subdomains, endpoints, queries, competitors)
  • Correlates findings across tools and sources to surface patterns humans might miss
  • Drafts research summaries and initial finding writeups for operator review
Findings log Evidence pack Mid-engagement readout
Step 05 · Drafting · 2 to 3 weeks

Claude drafts the handbook. A senior operator reviews every paragraph.

The drafting phase is where the compression shows up most. Claude writes the long-form chapters of the handbook, applies the intent markers from our template library, and produces a first complete draft in hours, not weeks. The operator then spends real time on the parts that matter: sharpening judgment calls, adding institutional context Claude does not know, cutting filler, and validating every claim. Nothing ships to the client without this review.

What Claude does
  • Runs the client intake + research through our template library (Nexmill)
  • Produces the first complete handbook draft (typically 60 to 120 pages)
  • Flags sections where its confidence is low so the operator can prioritize review
  • Runs self-critique and eval harness checks before handing to the operator
Handbook v0.8 draft Operator redlines Client preview call
Step 06 · Review · 1 week

Your team reads the draft, we meet, we revise.

You get the draft 7 days before the handover meeting so your team actually reads it. We run two sessions: one with the executive stakeholder focused on decisions and direction, and one with the engineering or operating team focused on detail and accuracy. We take all feedback in a shared doc, respond to every item in writing, and ship v1.0 within 5 business days.

What Claude does
  • Ingests all review feedback and drafts responses for each comment
  • Suggests specific edits to the handbook that would address each piece of feedback
  • Catches internal inconsistencies revealed by the feedback (rare, but they happen)
Review sessions x 2 Feedback response doc Handbook v1.0
Step 07 · Handover + post-care · 4 weeks

We hand over the artifact, run the walkthrough, and stay around for questions.

Handover is a 60 to 90 minute session where we walk your team through the handbook, answer questions, and agree on the post-engagement scope. For 30 days after handover, we answer questions async via the shared channel (cap: one hour/week, included in engagement). For retainer clients, the handbook becomes a living document and is updated monthly. Case study write-up (with your consent) is proposed at this step.

What Claude does
  • Generates a handover deck from the handbook for the walkthrough session
  • Produces a one-page executive summary for distribution inside your company
  • Drafts the case study (anonymized by default, named with written consent)
Final handbook (PDF + MD + Notion) Handover deck Executive one-pager 30 days async support
R4 What we need from you

Small list. Non-negotiable.

The engagement moves fast because we need very little from you. But the things we need, we need reliably. If any of these is impossible, let us know in step 1 and we will adjust or decline.

One decision maker. A single person who can approve scope, changes, and final sign-off. Not a committee. Not a forwarded email.
Access, promptly. The artifacts we need to do the work, granted within two business days of kickoff. Delays here delay the handbook.
Weekly 30 minutes. One scheduled sync per week, same day, same decision maker present. The rest is async.
Willingness to read. We are going to ship a long document. If your team does not have 2 hours to read it, the engagement will underdeliver regardless of our effort.
R5 The Claude rules

How we actually use the model.

Transparency about where Claude is on the critical path, and where it never is. These are the rules we do not break.

Claude always does

  • First drafts of every written deliverable
  • Summarization and correlation of research findings
  • Template application and long-form generation
  • Self-critique passes before operator review
  • Evaluation runs on every prompt version
  • Translation between technical and executive registers

Claude never does

  • Final sign-off on any client deliverable
  • Exploit execution against production systems
  • Decisions about engagement scope or pricing
  • Client-facing meetings without a human present
  • Code pushed to your production without human review
  • Anything not logged in the engagement tracker
Human-in-the-loop by default

Our pricing and our reputation are underwritten by human review. If you ever see a deliverable that looks like it came straight out of a model with no human judgment on it, that is a defect, and we want to know immediately. See the contact page to report a quality issue directly to the operator team.

R6 Communication

Async by default, predictable live.

Daily: one written update in your channel. End of working day, our time. Always sent. Never a surprise.
Weekly: one 30-minute live sync. Same day, same time. Decision maker present. Agenda in Slack 24h before.
Ad hoc: urgent questions answered within 4 hours during working hours. Escalation contact on the scope document.
Ready to start step 1

Book a 30 minute fit call.

We will send a three-question pre-brief the day before so our time together is spent on the actual problem.

Book a call