Skip to content
← THE LENS
START HERE

Where to begin.

The Lens is 46 playbooks across five stacks. Most readers don't need to read 46 things. They need to know which three to start with. This page is that map.

What's in The Lens

Two kinds of artefact.

  • Playbooks, written documents of 1,000 to 1,500 words each. The shape of an AI-leveraged marketing workflow, with the brief, the pipeline, the prompts, the eval harness, the failure modes, and either a real receipt (where Manual Focus has run this work) or an illustrative example (where it's the methodology distilled from the team's prior experience). Browse the full library →
  • Skills, runnable agent companions to 24 of the playbooks. SKILL.md files in the Anthropic Agent Skills spec, installable as a Claude Code plugin. The playbook explains the work, the skill is the work. Browse the skills →

New to prompting or AI skills? Read The Lens 101, a seven-minute primer on what a prompt is, what makes one good, what an agent skill is, and how to install and use them. Worth reading before the first playbook.

New to Cowork? Read the Cowork primer, the eight-minute setup guide for running Claude Code with your real tools loaded (email, calendar, CRM, transcripts) so the Lens playbooks act on the work you already do instead of generic examples.

Both are organised into five stacks based on what they help you ship. Pick the stack that matches the work in front of you this week.

BRAND

How the brand is positioned, the voice it writes in, the names and pillars that anchor it. 7 playbooks.
DEMAND

Channel-mix decisions, ambassador programmes, lifecycle journeys, the way demand actually gets created. 11 playbooks.
CONTENT

SEO clusters, eval-gated drafting, social, video, race-result automation, AI-news pipelines. 9 playbooks.
OPS

The connective tissue. Briefs, evals, guardrails, hiring, planning, launches, inventory, replacements. 9 playbooks.
PRODUCTIVITY

Daily briefings, email triage, meeting prep, call follow-up, pipeline rollups, OKR synthesis. The personal-assistant workflows that turn Cowork into a real second brain. 10 playbooks.

Four starting paths

The Lens is built to be entered from any direction, but four sequences cover most teams. Pick the one that matches the situation you're walking into.

PATH 1

You've inherited a brand and need an honest read

Most common starting situation. New CMO, fractional engagement, rebrand consideration, founder-led brand outgrowing its early positioning. Three plays in order.

  1. Positioning audit, running the brand's surface through a six-pass audit. The output is where the brand actually sits today, and where it should.
  2. Brand voice extraction, taking the audit's corpus and extracting the observable voice patterns. The output is a voice profile and scoring rubric for every future draft.
  3. Message house, converting the sharpened position into pillars, proof points and channel-mapped lines. The output is the messaging spine for the next 12 months.

TYPICAL ELAPSED TIME · 5 TO 10 WORKING DAYS

PATH 2

You need to ship better content, faster, in the brand's voice

You already know what the brand stands for. What you don't have is the production line that ships consistent work at the standard the brand needs. Three plays in order.

  1. Brand voice extraction. If you don't already have a voice profile in observable patterns, build one. Adjective-based voice guides won't work as input.
  2. Eval-gated drafting pipeline, the production line. Drafts score against four gates (voice, fact-grounded, originality, structural) before a human editor sees them.
  3. Social content factory, channel-native social production from a single weekly story brief. Stops cross-posting blog excerpts.

TYPICAL ELAPSED TIME · 1 TO 2 WEEKS TO INSTALL, ONGOING DAILY

PATH 3

You're building the marketing function's operating system

The function ships output but the system around it is ad-hoc. Briefs are chats, evals are missing, every piece needs senior review because the standard isn't codified. Three plays in order.

  1. Brief-to-ship pipeline, the structured brief template, the weekly triage, the production cadence, the post-mortem. Replaces ten tools and three weekly meetings.
  2. Evaluation frameworks, rubrics per output type. Deterministic where possible, LLM-as-judge where necessary. The standard becomes objective, not "the editor's gut."
  3. Brand guardrails as code, turning brand guidelines into a programmatic ruleset that runs against every piece before it ships. Catches drift before it ships, not after.

TYPICAL ELAPSED TIME · 2 TO 4 WEEKS TO INSTALL, COMPOUND BENEFIT FOR YEARS

PATH 4

You want your day back from email, meetings and admin

The marketing work is fine. The system around your inbox, calendar, Slack, CRM and meeting notes is eating five hours a day before any real work starts. The Productivity stack rebuilds that day as a Claude-assisted second brain. Three plays in order.

  1. Cowork primer, the setup guide. Install the Lens skills, connect your inbox, calendar, Slack, CRM and meeting transcript tool. Twenty to forty minutes one-off.
  2. Daily briefing pipeline, calendar plus inbox plus Slack plus CRM stitched into a five-minute morning brief. The first hour of your day becomes the first five minutes.
  3. Meeting prep stack and call follow-up loop, the bookends. Context pack before, CRM update plus follow-up draft plus tasks after, no manual write-up.

TYPICAL ELAPSED TIME · 1 TO 2 DAYS TO INSTALL, ONGOING DAILY

By situation

If the four paths don't match what you're walking in to, the table below points to the right starting playbook for specific situations. Each link drops you straight into the relevant entry.

If you're… Start with…
Naming a brand or product line Naming sprint
An endurance-sports brand and your content sounds generic Endurance brand voice
Trying to pick a tagline that won't embarrass you in two years Tagline system
Scaling past founder-only writing and need an institutional voice Founder and institutional voice
Trying to figure out which channels are actually working Attribution tear-down
Deciding how to allocate next quarter's spend Channel-mix simulator
Finding the buying moments the brand should be present for Category entry points
Building or rebuilding a lifecycle email programme Lifecycle journey builder
Planning campaigns around the race or event calendar Race-day demand pipeline
Restructuring an athlete or ambassador programme Ambassador programme
Building an event sponsorship programme that actually pays back Event sponsorship
Selling through bike shops, run stores or multisport retailers Retail partner programme
Designing a subscription or membership programme Subscription and membership
Building a coach-as-channel programme Direct-to-coach
Planning an SEO topic cluster SEO cluster generator
Getting press coverage without spray-and-pray Earned-media pitch
Producing race-day recaps within an hour of finish Race-result content engine
Planning a year of training-content for endurance audiences Training content engine
Setting up b-roll production for an endurance brand Segment b-roll production
Building a LinkedIn pipeline that explains AI releases AI studio news pipeline
Launching a new gear or apparel product Gear launch sequence
Cleaning up end-of-season inventory End-of-season inventory
Setting up a crash-replacement or wear-out programme Crash-replacement programme
Hiring for an AI-leveraged marketing team Hiring shape
Earning customer permission to use their race photos Customer content rights
Running a quarterly planning ritual that doesn't collapse Quarterly planning ritual

How the skills compose

24 of the playbooks ship as runnable skills for Claude Code, plus a setup-the-lens onboarding orchestrator and a video-script-to-heygen utility that renders avatar videos from playbook content, Cursor, Codex, Windsurf or any agent that honours the Agent Skills spec. Skills reference each other, with the output of one becoming the input of the next, so they compose into a real working cadence rather than 14 separate things.

Three composition patterns worth knowing.

Brand foundation

positioning-audit produces a positioning brief. brand-voice-extraction reads it for context. message-house builds pillars and channel lines. eval-gated-drafting uses the message house and voice rubric to gate every draft.

Endurance content engine

endurance-brand-voice extends the standard voice profile with sport-specific lexicon. seo-cluster-generator plans topic clusters around the brand's audience CEPs. eval-gated-drafting drafts each piece with voice and cluster context. social-content-factory repurposes each piece into channel-native social posts. race-result-content-engine ships race recaps fast.

Demand and lifecycle

category-entry-points identifies the buying triggers the brand should own. lifecycle-journey-builder uses the CEP map to define journey entry triggers and touchpoint copy. earned-media-pitch uses the CEP map to surface story angles for journalists. ai-studio-news-watch drafts LinkedIn posts about AI releases that compound the brand's CEP ownership.

Each skill's SKILL.md ends with a hand-off section naming the downstream skills it feeds. The composition is documented in the work, so you don't need to memorise it.

Install the plugin and set it up

The Lens Plugin in one line. 25 marketing workflows your AI agent runs inside Claude Code. Each skill pairs with a playbook on this site.

Step 1, install. Two commands in any Claude Code session install all 26 skills. 

/plugin marketplace add Scylark/manual-focus
/plugin install the-lens@manual-focus

Step 2, run setup. Ask Claude to set up The Lens. The onboarding skill asks what work is in front of you, recommends the Cowork plugins to install alongside, pulls your existing brand documentation from Notion or Drive or Figma into a structured .lens/ workspace, writes a task list for the foundational assets you do not have yet and recommends a starter playbook. Twenty to forty minutes end to end. 

set up the lens

Or run the slash command directly with /setup-the-lens.

Step 3, run the work. After setup, the agent recognises any prompt that matches a Lens skill. Asking something like "audit our positioning against the competitive set" surfaces the positioning-audit skill, pasting a race result URL with "draft the recap" surfaces the race-result-content-engine. The skill reads .lens/brand.json so you do not re-paste inputs every time.

Skills source at github.com/Scylark/manual-focus/tree/main/lens-skills ↗

Read the capability reference before you ship

The Lens is honest about what AI can and cannot reliably do this quarter. The capability reference grades every category, covering text, voice matching, original research, image (generic vs branded), video (generic vs branded vs named athletes), audio, browser automation and data analysis, and tells you which ones are production-ready, which need guardrails, which are experimental, and which are not-yet.

For endurance brands specifically, AI compounds time hard on analysis, drafting, orchestration and environmental b-roll variants. Practical shoots still matter for branded products and named athletes. The capability reference draws the line explicitly.

We re-grade quarterly. Updates ship as the model landscape moves.

What to do if…

You don't know which playbook applies

Ask Claude. With the plugin installed, describing your situation in plain language is enough, and the agent will pick the right skill. Without the plugin, the by-situation table above covers most starting points.

The playbook calls for a customer-call sprint and you don't have time

Some playbooks (positioning-audit, category-entry-points, retail-partner-programme) recommend a small customer interview set as input. They run without it but with lower confidence, and every playbook documents what running on degraded inputs actually costs you in the eval section.

You want to use a different model from the one the playbook names

Fine. Every playbook lists the models we've tested it against, and most workflows are model-agnostic with the right prompt scaffolding. The playbook's eval gates catch model drift directly. If a swap breaks output quality, the gates will tell you before the output ships.

You spot a gap in a playbook

Open a PR on github.com/Scylark/manual-focus or email the team. Real client work surfaces gaps faster than internal review, so we update the playbooks as soon as the gap is real.

You want to talk to Manual Focus instead

Some workflows benefit from a Manual Focus partner running them with you, especially the foundational ones (positioning, ambassador programmes, gear launches). Book a working session →

START NOW

Pick a path. Run one playbook this week.

The Lens compounds when used, not when read. The shortest path to value is to take one playbook into a real piece of work in front of you. Whichever you pick, the eval gates and failure-mode sections will catch you before you ship something the brand will regret.

Browse the library Browse the skills Read the capabilities