Project instructions applied · brand voice, ICP, current offer
NS
Audit this landing page and prioritise the fixes for the next build pass.
homepage-v3.png
@
Site Audit Operator·Landing Page AuditReviewed
Prioritised fix list
P1
Clarify the hero promiseImpact high · Effort low
cited
P1
Separate preview and sign-in pathsImpact high · Effort med
cited
P2
Cut the mobile explanation depthImpact med · Effort low
cited
P2
Put proof next to the primary actionImpact med · Effort med
cited
3 sources citedRun recordedSave to library
Site Audit Operator×Landing Page AuditMANUAL×Brand context×
Preview context3 sources · ~1.4k tokens
Refine, follow up, or change direction…
Attach context 3·@ operator · / playbook · Enter to send
Saved to library · trace attached
01 Backstage
The method is already attached.
You picked an operator and a playbook from the Studio, and the project has its knowledge pinned. You set them up once; the thread stays a normal conversation.
Operator
Site & App Audit Operator
The standing role. Its rules hold on every audit it runs.
The material the run works from, pinned to the project.
Voice, offer, who the page is for
Proof gaps found on earlier audits
02 The run
Simple where you type. Inspectable when it matters.
Each reply is backed by a run record you can open. Today a run is one model call with one operator and one playbook.
01Prompt snapshot
The exact text the model received, built from the operator, playbook, knowledge and your request.
02Context
Which sources were attached, and roughly how many tokens they took.
03Model and cost
The model that ran, tokens in and out, and what it cost.
04Checks
Whether the reply has every section the playbook asks for. Replies with no contract say Not checked.
03 The output
A deliverable, not a wall of chat.
The reply comes back in the playbook's sections, so it can be checked, compared with last time and handed on. Gaps are marked rather than filled with guesses.
Example output · Ranked issues5 of 5 sections
01The hero promise is vague. “Work smarter with AI” names no outcome. Lead with what the reader gets.
02No proof near the call to action.[PROOF NEEDED: a named customer result]
03Two competing buttons. Keep one primary action above the fold.
04 Kept
The next run starts richer.
Keep what was useful, and the next audit starts with it instead of a blank thread.
The outputsaved to the project library, with its run attached
The proof gaps you keep findingkept as knowledge and pinned to the project
The methodthe playbook keeps a version history as you improve it