Run AOA Full Assessment
run_aoa_full_assessmentAnalyze AI opportunity interview notes: screen tools, knowledge, skills and workflows; score impact and effort. Returns a draft record and report plan for host review, not files.
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"input": "<input>"
},
"name": "run_aoa_full_assessment"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| input | Yes | Interview notes or a full transcript, the business description, any earlier mini answers, an optional prior assessment record with its verification_needed list (JSON) and the requested scope, in the respondent's words with speaker labels. Send only information needed for this assessment; omit sensitive personal identifiers. Full assessment uses 16:9 slides. HOST REVIEW (static; each reply's host_review points here): Unreviewed draft. Compare every finding, quote and number with the current source; correct unsupported claims. metadata.client is the business name exactly as supplied, or Unknown when the source gives only a description. Add an opportunity for any task, repeated question, missing information, knowledge one person holds or wish in the source that the draft missed or folded into another, and merge opportunities that repeat one bottleneck. Recheck each impact and effort score, its quadrant and the matrix (a score of 7 or more is high: quick_wins is high impact with low effort, major_projects both high, fill_ins both low, thankless_tasks low impact with high effort, unscored when either score is unscored); move any record section (such as knowledge_domains, matrix, calculations, plan, discovery_gaps or skill_briefs) found outside record, or nested in another record section such as discovery, to the record's top level, and delete any key the item shapes do not list (such as analysis_only_reason). A task with no stated volume, consequence or process detail stays unscored. Preserve manual controls. Recheck each trial prerequisite: data permission is Not applicable unless customer details go into an AI tool (drafting for a particular customer from their message, invoice or record does), and a quote about approving output before it goes out is never prerequisite evidence. Remove any such quote; data permission then follows the rule just given, and trial approval is Unknown unless other evidence settles it. Keep every other prerequisite. A prerequisite marked Established or Not established without an exact source span becomes Unknown. Use only the values each field allows: prerequisite status Unknown, Established, Not established or Not applicable (Not applicable for data permission as above, for the baseline when the client does not do the task today, or for every prerequisite of an analysis-only item that changes nothing); opportunity status provisional, ready, blocked or deferred (analysis_only is a report_plan list, never a status: replace it with provisional); screen decision applicable, unnecessary or insufficient_evidence; knowledge modality systems, files, communications, spreadsheets, tacit or unknown. report_plan.recommended_choices holds recommendations for choices the source leaves open, not findings: keep each one, including the owner's approval. A recommended choice is never an opportunity: delete any opportunity that only restates one, such as an approval step for what goes out, from the record, matrix and report_plan. Every opportunity sits once across wins (at most four), later (at most three) and analysis_only (each analysis_only item an object with id and reason), and an item a client tool already handles is analysis_only. The helper is one role other than the owner (when the source names no one, propose one, such as a part-time assistant or student intern; never Unknown), with weekly hours the source states or To confirm. Drop any discovery gap that asks about a recommended choice. Each win gets a plan step that prepares on made-up inputs, and each win or later item whose types include skills has a skill brief with opportunity_id, name, trigger, inputs, decision_rules, output, and human_review; each verification claim states the fact to check. Strip any label, ID or comment from the quotes in calculations.assumptions. A verification_needed item the input already supplies as verified keeps its status, source_url and verified_on exactly as supplied. Check each other item on an official source and set status verified with its URL and date, or keep status unverified with its cost excluded. When report_plan is present, write the report deck and implementation guide copy yourself from the reviewed record and report_plan, following HOST REPORT COPY in the input parameter description; fill tool prices only from vendor pages checked today, compute any savings example yourself only from a quoted time for one task and a quoted weekly count (a weekly total such as two hours a week is not a time for one task) and recompute the cost subtotal. Ask the user for the client's logo; never invent one. Render with the pinned branding and layout in the input parameter description; verify asset hashes and run the self-inspection checklist on every page. Claim only actual downloads. Disclose corrections separately. STATIC HOST RENDERING SPECIFICATION. Do not send or regenerate this metadata in input or output. {"palette":{"canvas":"#000000","panel":"#1A1A1A","panel2":"#242424","accent":"#F06000","bright":"#FF7A1A","press":"#C04A00","head":"#FFFFFF","body":"#F0E0D0","muted":"#8090A0","rule":"rgba(240,224,208,.10)","rule2":"rgba(240,224,208,.22)"},"fonts":[{"family":"Montserrat","use":"display 800","url":"https://raw.githubusercontent.com/google/fonts/8b0a1d0f5983c89bc2b93f1b5fb55f9e252744b5/ofl/montserrat/Montserrat%5Bwght%5D.ttf","sha256":"0f7b311b2f3279e4eef9b2f968bcdbab6e28f4daeb1f049f4f278a902bcd82f7"},{"family":"Inter","use":"body 400/600","url":"https://raw.githubusercontent.com/google/fonts/0b58fb370093f9a9f4ff785d94405710b79de67c/ofl/inter/Inter%5Bopsz%2Cwght%5D.ttf","sha256":"29160a80ff49ddcab2c97711247e08b1fab27a484a329ce8b813d820dc559031"}],"license":"OFL.txt beside each font","logo_url":"https://api.cofr.io/mcp/fd73c601-5d90-4e3a-99b8-0ea5d5f853e4/icon?v=1790781628585","logo_sha256":"8987b040714d7893caef7320ae4eb5e2d45966dcb78e6eef6b2483cef0ae075f","geometry":{"full_report":"16:9 slides 1920x1080, one PDF page each","mini_report":"one A4 portrait page","discovery_guide":"A4 portrait","discovery_docx":"A4 portrait white page; #000000 or #1A1A1A heading bands, white Montserrat 800 uppercase, 3pt #F06000 rule; Inter #1A1A1A body; orange text #C04A00; editable styles, no page images or text boxes"},"footer":"Prepared with the AI Opportunity Assessment by AI Operator Academy.","fallback_wordmark":"AI OPERATOR ACADEMY","style":"Black full-bleed canvas, print backgrounds. Headings Montserrat 800 uppercase white (weight 800; file default 100). Eyebrows Montserrat 800 uppercase accent. Body Inter cream; labels mono uppercase muted in the system stack ui-monospace, 'SF Mono', Menlo, monospace (not embedded; needs no disclosure). Panels #1A1A1A, 1px rule border. Callouts: #F06000 fill, #000000 text, never white on orange. 72x3px accent rule under titles. Corners 0-4px.","assets":"Verify each SHA-256; embed both fonts in HTML and PDF. Never redraw, recolor or replace the logo. If the logo or a font fails, ask the user to attach it; else return the content in chat or, if the user accepts, use the wordmark in Montserrat 800 uppercase white over a 72x3px #F06000 rule (system-ui if Montserrat fails), disclosed in chat, not as the logo. Never deliver unbranded files silently.","self_inspection":"Inspect every page: Montserrat and Inter embedded or fallback disclosed; no overflow, clipping or overlap; text contrast at least 4.5:1, or 3:1 for large text (at least 24px or 18pt, or bold at least 18.66px or 14pt); footer on every page; claim only files created."} HOST REPORT COPY (static; the host writes this copy and the tool never returns it). After the review, write a briefing object (the report deck, which fills the blank slots of the AI Opportunity Assessment template) and a guide object (the implementation guide) yourself from the reviewed record and report_plan, then render them with the layout below. Shapes, with maximum characters: briefing {business_type 40; focus 40: the primary focus, such as Orders and weekly admin; pain 200: one or two sentences in the client's terms; outcome 200: what changes, report_plan.helper by name with weekly hours, and what the owner still does; wins: one per report_plan.wins ID in order {id, pain 70: the problem today, fix 110: the concrete result, tool 18: the solution's name, use 72: what it does, following any recommended choice, cost 16: such as $20/mo, Free, Included or To confirm, setup 16: such as 1 hour, saves: the owner's hours back a week as a number, or Measure first, or New capacity}; later: one per report_plan.later ID {id, text 70: what it makes possible, tool 40}; analysis_only: report_plan.analysis_only unchanged; plan: exactly 4 days when there is a win, none otherwise {task 130: a plain action naming who does it, tool 30: the tool used that day, opportunity_ids: wins only}; costs {checked: the date you read the vendor pages, YYYY-MM-DD; rows: up to 7 {item 70, kind tool (a new tool) or other (another new monthly cost the plan adds, such as the helper's pay; leave out what the client already pays), month: USD from the vendor page or To confirm, year: USD for an annual plan, otherwise twelve months, source_url: the vendor page}}; hourly_value: what one hour of the owner's time is worth in USD, only as the client stated it, otherwise null; financial {roi_cap 90, time_cap 90, cost_cap 90: one caption under each figure, such as its basis, the price-check date or what is still to confirm}; next_steps: exactly 2 {heading 40, detail 200}: first, start Day One with the implementation guide; second, what to count for four weeks before the review call}. guide {roles: 2-4 {who 40, does 160}; rollout: 2-6 rows in time order {when 30, set_up 160, who 40}; steps: an optional setup-only first step (empty id and outcome) when accounts must be set up first, then one per win in order repeating its id, its tool as the title and its fix as the outcome {id, title 40, outcome 140, when 40, setup: 1-6 of 160, routine_label 30, routine: 0-5 of 160, working_when 140}; skills: one per record.skill_briefs item with the same name {name 30, build_in 30, connects_to 80, instructions 600}}. Limits are maxima, not targets. Writing rules: plain words a busy owner reads in one pass, short sentences, you for the owner, everyday verbs; no hype words, dashes for rhythm or recap lines. Each fix is a concrete thing that now exists or happens; never save time, spend less time or reduce alone. Use report_plan.helper's name everywhere, and for a proposed role write your plus the role in lower case (your part-time assistant); never a helper or the team. Recommend instead of asking: state each report_plan.recommended_choices choice in the win's use, pick one option with a reason for anything still open, and never hand a choice back, list decisions or offer two options for one job. Undecided review of what goes out means the owner approves each item on the phone before it goes out; it is never a prerequisite or a plan day. Name the tools the client already has; a new tool is a category plus the features that matter, never AI tool; drafting uses the client's AI assistant. Each new tool gets a kind tool cost row priced from the vendor page checked today with its URL, at nonprofit or discount pricing where the client qualifies, or To confirm and left out of the total; a dollar amount on a solution card repeats its row's price. Saves: a number only from a savings example you compute when calculations.assumptions quotes, word for word from the source, both the time one task takes and how many happen a week (a weekly total such as two hours a week is not a per-task time) (net = count x (before - after - extra review) - upkeep, after and upkeep labelled assumptions; a negative net is added work, never a saving), otherwise Measure first; work not done today is New capacity, never hours back; never count the helper's hours. Figures: weekly time returned is the sum of the numeric saves (Measure first when none); total monthly tool cost is the sum of the priced kind tool rows ($0 with no new tool, To confirm when none is priced); monthly net ROI = weekly time returned x 52 / 12 x hourly_value - the sum of every row's month, shown only when hourly_value is set, the time is a number and every row is priced, otherwise Measure first; never estimate an hourly value. time_cap says four weeks of tracking will confirm the hours. Plan: plain actions with an owner; with fewer than four, add in order: sign up for the new tool (the helper), make the first example together (the owner and the helper), agree the weekly review time (the owner). Open trial prerequisites from the record become plan days in plain words (time the task, confirm data permission), never a client decision. Trial gate: no step uses customer details before data_processing_permission is Established; a routine that uses AI on live work starts only after the owner approves the trial. Guide: addressed to you; setup steps start with a verb and name who does them; roles describe each part once the plan runs; skill instructions come from the brief's inputs, decision rules and human review, add no facts, mark unsure items [CHECK] (at least one), end with the owner's review and never send or publish on their own. Knowledge in the owner's head: one-press voice notes of the owner alone, sent to the person who drafts. Keep IDs, quotes, evidence, research, legal detail, risk lectures and rules the client already knows out of the copy, and costs and hours out of the guide; unknowns say To confirm. Before rendering, check that every record opportunity appears once across wins, later and analysis_only, guide steps repeat the wins' tools and fixes, every priced row has a vendor URL and check date, and the three figures recompute. FULL ASSESSMENT LAYOUT (static, same rule): {"page_geometry":"16:9 report deck; A4 portrait implementation guide","layout_instructions":"Full assessment report deck: the AI Opportunity Assessment template, reproduced exactly with its blank slots filled from briefing: 16:9 slides at 1920x1080, one landscape PDF page per slide, in the shared style above on a full-bleed #000000 canvas (slides 6 and 9 on #1A1A1A). Keep each slide's order, fixed words, sizes and positions as given here; never restyle, rename or add slides. A slot without content is dropped, never padded or left blank. Safe area: padding 92px top, 120px left and right, 96px bottom; content starts 40px under the title. Header on slides 2-9: eyebrow in Montserrat 800 24px uppercase, letter spacing .22em, #F06000; the title 18px below in Montserrat 800 70px uppercase, line height 1.0, letter spacing .01em, #FFFFFF; no rule and no lead under it. Cards: #1A1A1A, 1px rule border, padding 34px 36px, square corners, no shadows. Body text Inter 400; labels mono uppercase #8090A0, letter spacing .14em, unless stated. Every slide, 46px from the bottom: the footer Prepared with the AI Opportunity Assessment by AI Operator Academy. in Inter 20px #8090A0 at left and the page number (01 / 09) in mono 24px #8090A0, letter spacing .08em, at right. (1) Cover: centred vertically, AI OPPORTUNITY in #FFFFFF over ASSESSMENT in #F06000, Montserrat 800 132px uppercase, line height .9; 30px below, Prepared for and the client name, or the business type when the client is Unknown (#FFFFFF, 600) in Inter 34px #F0E0D0. Pinned to the bottom of the safe area, a full-width row with a 1px rule top border and 30px top padding: three columns packed to the left, each only as wide as its text and 84px apart (never spread across the width): keys Assessment Date (the record date written out, such as October 26, 2026), Business Type and Primary Focus in mono 24px, values 10px below in Inter 27px #FFFFFF. The client's logo sits top right inside the safe area, 120px high on a #FFFFFF panel with 14px padding, only when the user attached it (ask once; never fetch, redraw or invent it; without it, no panel). (2) Executive Summary, eyebrow Where we are · where we’re going: a 1.25fr and 1fr grid with 32px gaps filling the slide. Left: two stacked equal cards, The Pain and The Outcome, each a Montserrat 800 30px uppercase #F06000 heading (letter spacing .16em) 18px above Inter 27px/1.5 #F0E0D0 text, centred vertically. Right: an orange #F06000 card twice the height of the card below it, with the weekly time returned in Montserrat 800 140px, line height .85, #000000, its unit hrs at 54px, and the fixed caption Hours you can reclaim every week 18px below in Inter 600 27px #000000; below, a card with Primary Focus in mono 24px over the focus in Montserrat 800 34px uppercase #FFFFFF, letter spacing .04em. Measure first sets on two lines at 100px. (3) Impact–Effort Matrix, eyebrow Where the value is: a 56px axis column beside a 2x2 grid of quadrant cards with 20px gaps, a 52px axis row below, 18px between. Axis labels in mono 24px #8090A0, letter spacing .16em: High Impact at top and Low Impact at bottom, turned to read upward; Low Effort, Effort and High Effort along the bottom. Each quadrant card (padding 30px 32px) has its name at top and its note at bottom, the notes exactly: Quick Wins (top left, a #F06000 card, name Montserrat 800 40px uppercase #000000, note Inter 600 24px #000000) High impact, low effort — this report focuses here.; Major Projects (top right) High impact, high effort — phase these in after the wins.; Fill-Ins (bottom left) Low impact, low effort — do them when time allows.; Ignore These (bottom right) Low impact, high effort — not worth the time now.; the other names in Montserrat 800 32px uppercase #8090A0 (letter spacing .04em) over 24px #8090A0 notes. Under each name, the numbers of the wins scored into that quadrant as 40px squares 10px apart in Montserrat 800 24px: #000000 on rgba(0,0,0,.18) in Quick Wins, #FFFFFF on #242424 elsewhere. Unscored wins go on one mono line under the grid, such as Not yet scored · 3, 4. Later items are not placed. (4) Quick Wins, eyebrow High impact, low effort: one row per win with 22px top and bottom padding and a 1px rule above each row and below the last, in columns 64px, 1fr, the arrow and 1.05fr with 28px gaps: the number 01 in Montserrat 800 34px #F06000, pain in Inter 500 27px #FFFFFF, → in 32px #F06000, fix in Inter 27px #F0E0D0. (5) Recommended Solutions, eyebrow The tool stack: a grid of 3 columns and 2 equal rows with 24px gaps filling the slide, one card per win in order (padding 28px 32px; the remaining positions stay empty): the number 01 in Montserrat 800 30px #F06000 and the tool in Montserrat 800 28px uppercase #FFFFFF (letter spacing .02em) on one baseline 14px apart; use 14px below in Inter 24px/1.4 #F0E0D0; at the card bottom, over a 1px rule with 18px above and below it, three rows 11px apart: Cost, Setup and Saves in mono 24px (letter spacing .04em) at left and the value in Montserrat 800 26px #FFFFFF at right without wrapping, Saves in #F06000 (a number reads 3 hrs a week). (6) Your 4-Day Quick Wins Plan, eyebrow Start this week, on #1A1A1A: four equal columns with 26px gaps filling the slide; each card #1A1A1A with a 1px rule border and padding 34px 32px: the number 1 to 4 in Montserrat 800 88px, line height .9, #F06000; Day One to Day Four 10px below in Montserrat 800 26px uppercase #FFFFFF (letter spacing .14em); the task 22px below in Inter 27px/1.45 #F0E0D0; at the bottom, over a 1px rule with 22px padding, one uppercase line that wraps only when it must: TOOL · in mono 24px #8090A0 (letter spacing .1em) and, on the same line, the tool in Montserrat 800 uppercase #F06000, such as TOOL · PHONE TIMER. (7) What Comes After Quick Wins, eyebrow The next phase: three columns 32px apart, no cards, centred vertically, one per later item (the remaining columns stay empty): the number 01 in Montserrat 800 96px, line height .85, #F06000; a 60x2px rule2 line with 26px above and below; text in Inter 500 42px/1.25 #FFFFFF; tool 18px below in mono 24px (letter spacing .1em). (8) Financial Impact, eyebrow The bottom line: a 1.4fr and 1fr grid of two rows with 32px gaps. Left, spanning both rows, an orange #F06000 card centred vertically: Monthly Net ROI in mono 24px #000000 at 70% opacity; the figure 14px below in Montserrat 800 168px, line height .85, #000000; roi_cap 20px below in Inter 600 27px #000000. Right, two cards centred vertically: Weekly Time Returned and Total Monthly Tool Cost in mono 24px, the figure 12px below in Montserrat 800 96px #FFFFFF with its unit (hrs, /mo) at 40px #8090A0, then time_cap and cost_cap 16px below in Inter 24px #8090A0. Measure first, or To confirm for an unpriced tool cost, sets on two lines at 120px on orange and 64px on the cards. (9) Your Next Steps, eyebrow Let’s get moving, on #1A1A1A: the two steps centred vertically, 30px apart, at most 1320px wide: the number in Montserrat 800 80px #F06000 in a 96px column, 34px gap; heading in Montserrat 800 42px uppercase #FFFFFF (letter spacing .02em); detail 12px below in Inter 27px/1.45 #F0E0D0, at most 980px wide; 14px under step 2, the button Schedule Your Review Call → on #F06000 in Montserrat 800 28px uppercase #000000 (letter spacing .08em), padding 24px 40px. All nine slides when there is a win and a later item; no slide 7 without a later item; with no win, slides 1, 2, 7 and 9 only; numbered in order (01 / 08 when one is dropped). Character limits in HOST REPORT COPY are maxima, not targets; shorten copy rather than shrinking type or clipping. Keep IDs, evidence, quotes, research and verification detail off the slides and out of the guide. Use Montserrat, Inter and the authentic logo only after verifying the supplied asset hashes. Never redraw, recolor or replace the logo. If the logo or a font cannot be fetched or verified, ask the user to attach the verified files. Without them, return the reviewed content in chat and explain the limitation, or, if the user accepts, use the fallback: the wordmark AI OPERATOR ACADEMY in Montserrat 800 uppercase #FFFFFF over a 72x3px #F06000 rule (system-ui if Montserrat is unverified), disclosed in chat and never presented as the logo. Never deliver unbranded files silently. Keep test/debug disclosures outside the customer-facing files.","implementation_guide":"From the guide object: an A4 portrait reading document, HTML and PDF, on the black canvas; one column at most 1000px wide with a 16px side gutter, Inter 18px/1.6 #F0E0D0. Top bar in Montserrat 800 13px uppercase #8090A0: AI Opportunity Assessment · Implementation guide, with a #F06000 link back to the report PDF. Title Implementation guide (Montserrat 800 44px uppercase #FFFFFF), client and date in mono uppercase #8090A0, one intro line. Then: Who does what (roles as a list, the who in #FFFFFF 600); The rollout (a table When, Set up, Who); each step as Step N: title (Montserrat 800 32px uppercase #FFFFFF under a 3px #F06000 top rule), its outcome in a #1A1A1A callout with a 3px #F06000 left border and #FFFFFF text, a Set up (when) subheading (Montserrat 800 20px uppercase #F06000) over a numbered list, the routine_label subheading over a numbered list, and It's working when in #FFFFFF 600 before the check; last, Build a skill: three numbered steps (create it in the client's AI assistant and paste the starter instructions; test it on one made-up example and fix what's off; share it with report_plan.helper), a table Skill, Build in, Connects to, then each skill's name over its instructions in a #1A1A1A mono block. Table headers Montserrat 800 12px uppercase #F06000 on #1A1A1A, rows 15px with rule dividers; list numbers bold #F06000, continuing after an instruction block. The HTML adds a contents list; the PDF omits it. Print A4 with backgrounds; never split a callout, numbered list, table row or instruction block across pages or leave a heading at a page bottom. Footer text on every page.","interactive_html":"Optional standalone HTML of the deck when the host can build it: the same slides with a labelled left sidebar (220px, #1A1A1A, 1px rule2 right border, items Inter 16px #F0E0D0; the current slide gets a #242424 fill, a 3px #F06000 left border, #FFFFFF text and a square #F06000 number chip with #000000 text), a 72px icon rail from 701 to 1099px wide, a compact top strip at 700px and below, keyboard navigation with a 3px #F06000 focus ring at 4px offset, matrix numbers that open detail dialogs (#1A1A1A, 5px #F06000 top border: the tool, its fix, the problem today and its saves) and return focus on close, a checklist on the four plan days that is a reading aid with no storage, a link to the guide PDF and the exact PDF bytes embedded as a download. No slide transitions. Embed both fonts. No external requests, remote scripts, browser storage, accounts or comments. Minimum delivery: the deck PDF and the guide PDF, each with its HTML.","self_inspection":"Self-inspection before delivery: both font SHA-256s verified and Montserrat and Inter embedded in HTML and PDF, with display text at weight 800, or the fallback disclosed; logo verified bytes or the disclosed wordmark; the client logo only as attached; full-bleed canvas on every PDF page (backgrounds printed); no white or cream text on #F06000; every slide 16:9 1920x1080 and laid out as the template above, with no blank slot, no overflow, clipping or overlap, all content inside the safe area clear of the footer and page number, and nothing below the minimum sizes; footer on every slide and guide page; PDF page count equals slide count; win numbers match across the matrix, quick wins and solution cards, and the guide's steps repeat each win's tool and fix; every price has a vendor URL and check date in verification_needed, card prices repeat their rows and the tool total recomputes; saves follow the savings example you computed from calculations.assumptions or say Measure first or New capacity; the summary's hours equal Weekly Time Returned; the ROI is a figure only from a stated hourly value with every cost priced; every opportunity appears once as a win, later item or analysis_only item; unscored wins listed as Not yet scored; no IDs, quotes, sources or legal detail on the slides or in the guide, and no costs or hours in the guide; the guide's last step links back to the report; HTML and PDF carry the same content; claim only files actually created."} |
Output Schema
| Name | Required | Description | Default |
|---|---|---|---|
| scope | No | Text. One of: "full", "report", "discovery_analysis", "tools", "knowledge", "skills", "workflows", "prioritize", "handoff". | |
| action | No | Text. out_of_scope first for external actions or fabrication; needs_evidence when nothing to analyze was supplied; otherwise assessment, even for sparse evidence. One of: "assessment", "needs_evidence", "out_of_scope". | |
| record | No | An object. Assessment record, schema 1.0.0. Only IDs, scores, quadrants, each opportunity's trial_prerequisites statuses, the matrix and savings arithmetic are required, so a long reply is not rejected; still fill every key the method calls for, including trial_prerequisites evidence, evidence_ids and discovery_gaps. Out-of-scope sections keep pasted values or stay empty. The record closes only after workflow_briefs: matrix, calculations, plan and discovery_gaps sit inside it, never at the top level. | |
| message | No | Text. | |
| delivery | No | An object. AUTHORITATIVE STATIC RENDERING METADATA (not generated report content). Verify downloaded asset bytes against pinned SHA-256 hashes before embedding and run the self-inspection checklist on every exported page. Use the black canvas with Montserrat and Inter; no silent font, logo, or white-template substitutions. STATIC HOST RENDERING SPECIFICATION. Do not send or regenerate this metadata in input or output. {"palette":{"canvas":"#000000","panel":"#1A1A1A","panel2":"#242424","accent":"#F06000","bright":"#FF7A1A","press":"#C04A00","head":"#FFFFFF","body":"#F0E0D0","muted":"#8090A0","rule":"rgba(240,224,208,.10)","rule2":"rgba(240,224,208,.22)"},"fonts":[{"family":"Montserrat","use":"display 800","url":"https://raw.githubusercontent.com/google/fonts/8b0a1d0f5983c89bc2b93f1b5fb55f9e252744b5/ofl/montserrat/Montserrat%5Bwght%5D.ttf","sha256":"0f7b311b2f3279e4eef9b2f968bcdbab6e28f4daeb1f049f4f278a902bcd82f7"},{"family":"Inter","use":"body 400/600","url":"https://raw.githubusercontent.com/google/fonts/0b58fb370093f9a9f4ff785d94405710b79de67c/ofl/inter/Inter%5Bopsz%2Cwght%5D.ttf","sha256":"29160a80ff49ddcab2c97711247e08b1fab27a484a329ce8b813d820dc559031"}],"license":"OFL.txt beside each font","logo_url":"https://api.cofr.io/mcp/fd73c601-5d90-4e3a-99b8-0ea5d5f853e4/icon?v=1790781628585","logo_sha256":"8987b040714d7893caef7320ae4eb5e2d45966dcb78e6eef6b2483cef0ae075f","geometry":{"full_report":"16:9 slides 1920x1080, one PDF page each","mini_report":"one A4 portrait page","discovery_guide":"A4 portrait","discovery_docx":"A4 portrait white page; #000000 or #1A1A1A heading bands, white Montserrat 800 uppercase, 3pt #F06000 rule; Inter #1A1A1A body; orange text #C04A00; editable styles, no page images or text boxes"},"footer":"Prepared with the AI Opportunity Assessment by AI Operator Academy.","fallback_wordmark":"AI OPERATOR ACADEMY","style":"Black full-bleed canvas, print backgrounds. Headings Montserrat 800 uppercase white (weight 800; file default 100). Eyebrows Montserrat 800 uppercase accent. Body Inter cream; labels mono uppercase muted in the system stack ui-monospace, 'SF Mono', Menlo, monospace (not embedded; needs no disclosure). Panels #1A1A1A, 1px rule border. Callouts: #F06000 fill, #000000 text, never white on orange. 72x3px accent rule under titles. Corners 0-4px.","assets":"Verify each SHA-256; embed both fonts in HTML and PDF. Never redraw, recolor or replace the logo. If the logo or a font fails, ask the user to attach it; else return the content in chat or, if the user accepts, use the wordmark in Montserrat 800 uppercase white over a 72x3px #F06000 rule (system-ui if Montserrat fails), disclosed in chat, not as the logo. Never deliver unbranded files silently.","self_inspection":"Inspect every page: Montserrat and Inter embedded or fallback disclosed; no overflow, clipping or overlap; text contrast at least 4.5:1, or 3:1 for large text (at least 24px or 18pt, or bold at least 18.66px or 14pt); footer on every page; claim only files created."} HOST REPORT COPY (static; the host writes this copy and the tool never returns it). After the review, write a briefing object (the report deck, which fills the blank slots of the AI Opportunity Assessment template) and a guide object (the implementation guide) yourself from the reviewed record and report_plan, then render them with the layout below. Shapes, with maximum characters: briefing {business_type 40; focus 40: the primary focus, such as Orders and weekly admin; pain 200: one or two sentences in the client's terms; outcome 200: what changes, report_plan.helper by name with weekly hours, and what the owner still does; wins: one per report_plan.wins ID in order {id, pain 70: the problem today, fix 110: the concrete result, tool 18: the solution's name, use 72: what it does, following any recommended choice, cost 16: such as $20/mo, Free, Included or To confirm, setup 16: such as 1 hour, saves: the owner's hours back a week as a number, or Measure first, or New capacity}; later: one per report_plan.later ID {id, text 70: what it makes possible, tool 40}; analysis_only: report_plan.analysis_only unchanged; plan: exactly 4 days when there is a win, none otherwise {task 130: a plain action naming who does it, tool 30: the tool used that day, opportunity_ids: wins only}; costs {checked: the date you read the vendor pages, YYYY-MM-DD; rows: up to 7 {item 70, kind tool (a new tool) or other (another new monthly cost the plan adds, such as the helper's pay; leave out what the client already pays), month: USD from the vendor page or To confirm, year: USD for an annual plan, otherwise twelve months, source_url: the vendor page}}; hourly_value: what one hour of the owner's time is worth in USD, only as the client stated it, otherwise null; financial {roi_cap 90, time_cap 90, cost_cap 90: one caption under each figure, such as its basis, the price-check date or what is still to confirm}; next_steps: exactly 2 {heading 40, detail 200}: first, start Day One with the implementation guide; second, what to count for four weeks before the review call}. guide {roles: 2-4 {who 40, does 160}; rollout: 2-6 rows in time order {when 30, set_up 160, who 40}; steps: an optional setup-only first step (empty id and outcome) when accounts must be set up first, then one per win in order repeating its id, its tool as the title and its fix as the outcome {id, title 40, outcome 140, when 40, setup: 1-6 of 160, routine_label 30, routine: 0-5 of 160, working_when 140}; skills: one per record.skill_briefs item with the same name {name 30, build_in 30, connects_to 80, instructions 600}}. Limits are maxima, not targets. Writing rules: plain words a busy owner reads in one pass, short sentences, you for the owner, everyday verbs; no hype words, dashes for rhythm or recap lines. Each fix is a concrete thing that now exists or happens; never save time, spend less time or reduce alone. Use report_plan.helper's name everywhere, and for a proposed role write your plus the role in lower case (your part-time assistant); never a helper or the team. Recommend instead of asking: state each report_plan.recommended_choices choice in the win's use, pick one option with a reason for anything still open, and never hand a choice back, list decisions or offer two options for one job. Undecided review of what goes out means the owner approves each item on the phone before it goes out; it is never a prerequisite or a plan day. Name the tools the client already has; a new tool is a category plus the features that matter, never AI tool; drafting uses the client's AI assistant. Each new tool gets a kind tool cost row priced from the vendor page checked today with its URL, at nonprofit or discount pricing where the client qualifies, or To confirm and left out of the total; a dollar amount on a solution card repeats its row's price. Saves: a number only from a savings example you compute when calculations.assumptions quotes, word for word from the source, both the time one task takes and how many happen a week (a weekly total such as two hours a week is not a per-task time) (net = count x (before - after - extra review) - upkeep, after and upkeep labelled assumptions; a negative net is added work, never a saving), otherwise Measure first; work not done today is New capacity, never hours back; never count the helper's hours. Figures: weekly time returned is the sum of the numeric saves (Measure first when none); total monthly tool cost is the sum of the priced kind tool rows ($0 with no new tool, To confirm when none is priced); monthly net ROI = weekly time returned x 52 / 12 x hourly_value - the sum of every row's month, shown only when hourly_value is set, the time is a number and every row is priced, otherwise Measure first; never estimate an hourly value. time_cap says four weeks of tracking will confirm the hours. Plan: plain actions with an owner; with fewer than four, add in order: sign up for the new tool (the helper), make the first example together (the owner and the helper), agree the weekly review time (the owner). Open trial prerequisites from the record become plan days in plain words (time the task, confirm data permission), never a client decision. Trial gate: no step uses customer details before data_processing_permission is Established; a routine that uses AI on live work starts only after the owner approves the trial. Guide: addressed to you; setup steps start with a verb and name who does them; roles describe each part once the plan runs; skill instructions come from the brief's inputs, decision rules and human review, add no facts, mark unsure items [CHECK] (at least one), end with the owner's review and never send or publish on their own. Knowledge in the owner's head: one-press voice notes of the owner alone, sent to the person who drafts. Keep IDs, quotes, evidence, research, legal detail, risk lectures and rules the client already knows out of the copy, and costs and hours out of the guide; unknowns say To confirm. Before rendering, check that every record opportunity appears once across wins, later and analysis_only, guide steps repeat the wins' tools and fixes, every priced row has a vendor URL and check date, and the three figures recompute. FULL ASSESSMENT LAYOUT (static, same rule): {"page_geometry":"16:9 report deck; A4 portrait implementation guide","layout_instructions":"Full assessment report deck: the AI Opportunity Assessment template, reproduced exactly with its blank slots filled from briefing: 16:9 slides at 1920x1080, one landscape PDF page per slide, in the shared style above on a full-bleed #000000 canvas (slides 6 and 9 on #1A1A1A). Keep each slide's order, fixed words, sizes and positions as given here; never restyle, rename or add slides. A slot without content is dropped, never padded or left blank. Safe area: padding 92px top, 120px left and right, 96px bottom; content starts 40px under the title. Header on slides 2-9: eyebrow in Montserrat 800 24px uppercase, letter spacing .22em, #F06000; the title 18px below in Montserrat 800 70px uppercase, line height 1.0, letter spacing .01em, #FFFFFF; no rule and no lead under it. Cards: #1A1A1A, 1px rule border, padding 34px 36px, square corners, no shadows. Body text Inter 400; labels mono uppercase #8090A0, letter spacing .14em, unless stated. Every slide, 46px from the bottom: the footer Prepared with the AI Opportunity Assessment by AI Operator Academy. in Inter 20px #8090A0 at left and the page number (01 / 09) in mono 24px #8090A0, letter spacing .08em, at right. (1) Cover: centred vertically, AI OPPORTUNITY in #FFFFFF over ASSESSMENT in #F06000, Montserrat 800 132px uppercase, line height .9; 30px below, Prepared for and the client name, or the business type when the client is Unknown (#FFFFFF, 600) in Inter 34px #F0E0D0. Pinned to the bottom of the safe area, a full-width row with a 1px rule top border and 30px top padding: three columns packed to the left, each only as wide as its text and 84px apart (never spread across the width): keys Assessment Date (the record date written out, such as October 26, 2026), Business Type and Primary Focus in mono 24px, values 10px below in Inter 27px #FFFFFF. The client's logo sits top right inside the safe area, 120px high on a #FFFFFF panel with 14px padding, only when the user attached it (ask once; never fetch, redraw or invent it; without it, no panel). (2) Executive Summary, eyebrow Where we are · where we’re going: a 1.25fr and 1fr grid with 32px gaps filling the slide. Left: two stacked equal cards, The Pain and The Outcome, each a Montserrat 800 30px uppercase #F06000 heading (letter spacing .16em) 18px above Inter 27px/1.5 #F0E0D0 text, centred vertically. Right: an orange #F06000 card twice the height of the card below it, with the weekly time returned in Montserrat 800 140px, line height .85, #000000, its unit hrs at 54px, and the fixed caption Hours you can reclaim every week 18px below in Inter 600 27px #000000; below, a card with Primary Focus in mono 24px over the focus in Montserrat 800 34px uppercase #FFFFFF, letter spacing .04em. Measure first sets on two lines at 100px. (3) Impact–Effort Matrix, eyebrow Where the value is: a 56px axis column beside a 2x2 grid of quadrant cards with 20px gaps, a 52px axis row below, 18px between. Axis labels in mono 24px #8090A0, letter spacing .16em: High Impact at top and Low Impact at bottom, turned to read upward; Low Effort, Effort and High Effort along the bottom. Each quadrant card (padding 30px 32px) has its name at top and its note at bottom, the notes exactly: Quick Wins (top left, a #F06000 card, name Montserrat 800 40px uppercase #000000, note Inter 600 24px #000000) High impact, low effort — this report focuses here.; Major Projects (top right) High impact, high effort — phase these in after the wins.; Fill-Ins (bottom left) Low impact, low effort — do them when time allows.; Ignore These (bottom right) Low impact, high effort — not worth the time now.; the other names in Montserrat 800 32px uppercase #8090A0 (letter spacing .04em) over 24px #8090A0 notes. Under each name, the numbers of the wins scored into that quadrant as 40px squares 10px apart in Montserrat 800 24px: #000000 on rgba(0,0,0,.18) in Quick Wins, #FFFFFF on #242424 elsewhere. Unscored wins go on one mono line under the grid, such as Not yet scored · 3, 4. Later items are not placed. (4) Quick Wins, eyebrow High impact, low effort: one row per win with 22px top and bottom padding and a 1px rule above each row and below the last, in columns 64px, 1fr, the arrow and 1.05fr with 28px gaps: the number 01 in Montserrat 800 34px #F06000, pain in Inter 500 27px #FFFFFF, → in 32px #F06000, fix in Inter 27px #F0E0D0. (5) Recommended Solutions, eyebrow The tool stack: a grid of 3 columns and 2 equal rows with 24px gaps filling the slide, one card per win in order (padding 28px 32px; the remaining positions stay empty): the number 01 in Montserrat 800 30px #F06000 and the tool in Montserrat 800 28px uppercase #FFFFFF (letter spacing .02em) on one baseline 14px apart; use 14px below in Inter 24px/1.4 #F0E0D0; at the card bottom, over a 1px rule with 18px above and below it, three rows 11px apart: Cost, Setup and Saves in mono 24px (letter spacing .04em) at left and the value in Montserrat 800 26px #FFFFFF at right without wrapping, Saves in #F06000 (a number reads 3 hrs a week). (6) Your 4-Day Quick Wins Plan, eyebrow Start this week, on #1A1A1A: four equal columns with 26px gaps filling the slide; each card #1A1A1A with a 1px rule border and padding 34px 32px: the number 1 to 4 in Montserrat 800 88px, line height .9, #F06000; Day One to Day Four 10px below in Montserrat 800 26px uppercase #FFFFFF (letter spacing .14em); the task 22px below in Inter 27px/1.45 #F0E0D0; at the bottom, over a 1px rule with 22px padding, one uppercase line that wraps only when it must: TOOL · in mono 24px #8090A0 (letter spacing .1em) and, on the same line, the tool in Montserrat 800 uppercase #F06000, such as TOOL · PHONE TIMER. (7) What Comes After Quick Wins, eyebrow The next phase: three columns 32px apart, no cards, centred vertically, one per later item (the remaining columns stay empty): the number 01 in Montserrat 800 96px, line height .85, #F06000; a 60x2px rule2 line with 26px above and below; text in Inter 500 42px/1.25 #FFFFFF; tool 18px below in mono 24px (letter spacing .1em). (8) Financial Impact, eyebrow The bottom line: a 1.4fr and 1fr grid of two rows with 32px gaps. Left, spanning both rows, an orange #F06000 card centred vertically: Monthly Net ROI in mono 24px #000000 at 70% opacity; the figure 14px below in Montserrat 800 168px, line height .85, #000000; roi_cap 20px below in Inter 600 27px #000000. Right, two cards centred vertically: Weekly Time Returned and Total Monthly Tool Cost in mono 24px, the figure 12px below in Montserrat 800 96px #FFFFFF with its unit (hrs, /mo) at 40px #8090A0, then time_cap and cost_cap 16px below in Inter 24px #8090A0. Measure first, or To confirm for an unpriced tool cost, sets on two lines at 120px on orange and 64px on the cards. (9) Your Next Steps, eyebrow Let’s get moving, on #1A1A1A: the two steps centred vertically, 30px apart, at most 1320px wide: the number in Montserrat 800 80px #F06000 in a 96px column, 34px gap; heading in Montserrat 800 42px uppercase #FFFFFF (letter spacing .02em); detail 12px below in Inter 27px/1.45 #F0E0D0, at most 980px wide; 14px under step 2, the button Schedule Your Review Call → on #F06000 in Montserrat 800 28px uppercase #000000 (letter spacing .08em), padding 24px 40px. All nine slides when there is a win and a later item; no slide 7 without a later item; with no win, slides 1, 2, 7 and 9 only; numbered in order (01 / 08 when one is dropped). Character limits in HOST REPORT COPY are maxima, not targets; shorten copy rather than shrinking type or clipping. Keep IDs, evidence, quotes, research and verification detail off the slides and out of the guide. Use Montserrat, Inter and the authentic logo only after verifying the supplied asset hashes. Never redraw, recolor or replace the logo. If the logo or a font cannot be fetched or verified, ask the user to attach the verified files. Without them, return the reviewed content in chat and explain the limitation, or, if the user accepts, use the fallback: the wordmark AI OPERATOR ACADEMY in Montserrat 800 uppercase #FFFFFF over a 72x3px #F06000 rule (system-ui if Montserrat is unverified), disclosed in chat and never presented as the logo. Never deliver unbranded files silently. Keep test/debug disclosures outside the customer-facing files.","implementation_guide":"From the guide object: an A4 portrait reading document, HTML and PDF, on the black canvas; one column at most 1000px wide with a 16px side gutter, Inter 18px/1.6 #F0E0D0. Top bar in Montserrat 800 13px uppercase #8090A0: AI Opportunity Assessment · Implementation guide, with a #F06000 link back to the report PDF. Title Implementation guide (Montserrat 800 44px uppercase #FFFFFF), client and date in mono uppercase #8090A0, one intro line. Then: Who does what (roles as a list, the who in #FFFFFF 600); The rollout (a table When, Set up, Who); each step as Step N: title (Montserrat 800 32px uppercase #FFFFFF under a 3px #F06000 top rule), its outcome in a #1A1A1A callout with a 3px #F06000 left border and #FFFFFF text, a Set up (when) subheading (Montserrat 800 20px uppercase #F06000) over a numbered list, the routine_label subheading over a numbered list, and It's working when in #FFFFFF 600 before the check; last, Build a skill: three numbered steps (create it in the client's AI assistant and paste the starter instructions; test it on one made-up example and fix what's off; share it with report_plan.helper), a table Skill, Build in, Connects to, then each skill's name over its instructions in a #1A1A1A mono block. Table headers Montserrat 800 12px uppercase #F06000 on #1A1A1A, rows 15px with rule dividers; list numbers bold #F06000, continuing after an instruction block. The HTML adds a contents list; the PDF omits it. Print A4 with backgrounds; never split a callout, numbered list, table row or instruction block across pages or leave a heading at a page bottom. Footer text on every page.","interactive_html":"Optional standalone HTML of the deck when the host can build it: the same slides with a labelled left sidebar (220px, #1A1A1A, 1px rule2 right border, items Inter 16px #F0E0D0; the current slide gets a #242424 fill, a 3px #F06000 left border, #FFFFFF text and a square #F06000 number chip with #000000 text), a 72px icon rail from 701 to 1099px wide, a compact top strip at 700px and below, keyboard navigation with a 3px #F06000 focus ring at 4px offset, matrix numbers that open detail dialogs (#1A1A1A, 5px #F06000 top border: the tool, its fix, the problem today and its saves) and return focus on close, a checklist on the four plan days that is a reading aid with no storage, a link to the guide PDF and the exact PDF bytes embedded as a download. No slide transitions. Embed both fonts. No external requests, remote scripts, browser storage, accounts or comments. Minimum delivery: the deck PDF and the guide PDF, each with its HTML.","self_inspection":"Self-inspection before delivery: both font SHA-256s verified and Montserrat and Inter embedded in HTML and PDF, with display text at weight 800, or the fallback disclosed; logo verified bytes or the disclosed wordmark; the client logo only as attached; full-bleed canvas on every PDF page (backgrounds printed); no white or cream text on #F06000; every slide 16:9 1920x1080 and laid out as the template above, with no blank slot, no overflow, clipping or overlap, all content inside the safe area clear of the footer and page number, and nothing below the minimum sizes; footer on every slide and guide page; PDF page count equals slide count; win numbers match across the matrix, quick wins and solution cards, and the guide's steps repeat each win's tool and fix; every price has a vendor URL and check date in verification_needed, card prices repeat their rows and the tool total recomputes; saves follow the savings example you computed from calculations.assumptions or say Measure first or New capacity; the summary's hours equal Weekly Time Returned; the ROI is a figure only from a stated hourly value with every cost priced; every opportunity appears once as a win, later item or analysis_only item; unscored wins listed as Not yet scored; no IDs, quotes, sources or legal detail on the slides or in the guide, and no costs or hours in the guide; the guide's last step links back to the report; HTML and PDF carry the same content; claim only files actually created."} | |
| artifacts | No | A list. Always empty: this tool does not generate files. | |
| host_review | No | An object. | |
| report_plan | No | An object. Analyst choices the host's report and guide follow. Write it before the record, with the opportunity IDs the record then defines. Scope full or report only; for every other scope omit the whole object. No client copy. | |
| handoff_brief | No | An object. Implementation brief. Scope handoff only: fill every field. For every other scope omit the whole object. | |
| contract_version | No | Text. Expected value: "0.2.26". | |
| verification_needed | No | A list. Every current tool capability, plan or price claim. The host verifies or keeps it provisional. |