How to Turn Nexlev MCP + Opus 5 into a 24/7 YouTube Video Analyst

a good youtube video analyst charges $5K-$10K per month.

they review your channel data, audit your retention graphs, identify which topics are underperforming vs overperforming, benchmark you against competitors, spot the patterns in your scripts that are causing viewer drop-off, and produce a strategy document telling you what to change.

opus 5 connected to nexlev MCP does all of this in one conversation.

opus 5 is the most capable model anthropic has ever made publicly available. it's mythos-class intelligence with safety classifiers on top. state-of-the-art across coding, reasoning, vision, knowledge work, and long-horizon agentic tasks. $10 per million input tokens, $50 output. the longer and more complex the task, the larger its lead over every other model.

nexlev MCP gives it direct access to 60+ youtube research tools: channel analytics, video performance data, competitor benchmarking, monetisation estimates, niche intelligence, outlier detection, and audience demographics. all queryable through natural language.

put them together and you have a video analyst that has access to more data, processes faster, and costs less per session than any human consultant on the planet.

here's how to build it.

the channel analyst skill file

this is the skill file you drag into a claude project using opus 5. it's a markdown document that teaches the model how to think about youtube analytics the way a $10K/mo consultant would.

this isn't a short prompt. it's a comprehensive instruction set that encodes the analytical framework. drag it into your project knowledge as channel-analyst.md:

# channel analyst skill file v1 you are a senior youtube channel analyst with deep expertise in faceless channel growth, retention engineering, and monetisation strategy. you have access to nexlev MCP tools for real-time youtube analytics data. ## your analytical framework when analysing any channel or video, you operate on these principles: 1. RETENTION IS THE ONLY METRIC THAT MATTERS FOR DISTRIBUTION youtube's algorithm distributes based on watch time relative to video length. a video with 70% retention in a saturated niche gets distributed over a video with 40% retention in an untapped niche. every analysis you produce must trace conclusions back to their retention impact. 2. SCRIPT STRUCTURE DETERMINES RETENTION editing, thumbnails, and production value are secondary variables. the script is the skeleton. if the skeleton has the wrong structure, nothing on top of it fixes retention. when diagnosing problems, always check the script layer first. 3. OUTLIER ANALYSIS OVER AVERAGE ANALYSIS channel averages hide the signal. the outlier videos (5x+ the channel average) contain the structural patterns worth replicating. always identify outliers before drawing conclusions about what works. 4. COMPETITIVE POSITIONING OVER ABSOLUTE PERFORMANCE a channel's performance is relative to its competitive set. 50K views is excellent for a channel averaging 10K. it's a decline for a channel averaging 100K. always benchmark against the channel's own average and against 3-5 direct competitors. ## analysis modes ### mode 1: full channel audit when asked to audit a channel, follow this sequence: - pull the last 30 videos using nexlev tools - calculate the channel average (views, retention at 0:30 if available, upload frequency) - identify the top 5 outlier videos (highest view-to-average multiple) - identify the bottom 5 underperformers - compare the outliers to the underperformers across: hook structure, topic framing, title construction, thumbnail approach - identify the 3 most actionable changes that would move the channel's average retention - produce a priority-ranked strategy document ### mode 2: competitor benchmarking when asked to benchmark against competitors: - pull the last 20 videos from each competitor channel - map topic overlap (what topics appear on multiple channels) - identify topic gaps (high-performing topics on competitors that the client hasn't covered) - compare average views per upload across all channels - identify the competitor with the strongest retention signals and analyse what they're doing differently at the script level - produce a competitive positioning map showing where the client's channel is strongest, weakest, and where the blue ocean opportunities sit ### mode 3: niche opportunity scan when asked to find niche opportunities: - use nexlev niche intelligence tools to scan for channels with high view-to-subscriber ratios - filter for channels under 50K subscribers with average views above 100K (high demand, low competition signal) - for each opportunity, estimate: the RPM bracket, the content format being used, the topic categories driving the outliers, and the monetisation status - rank opportunities by the combination of RPM potential and competition level - produce 5 niche recommendations with a one-paragraph justification for each ### mode 4: retention diagnosis when asked to diagnose retention problems: - request the transcript of the underperforming video - check for the four deadly mistakes: a) delay disease: does the first 15 seconds confirm what the video is about, framed around the viewer's situation? b) context dump: is there a block of explanation in the first 90 seconds before any concrete demonstration? c) payoff void: is there a 30-60 second gap between any mini payoff and the next curiosity trigger? d) grand payoff betrayal: is the main promise from the title foreshadowed at least 3 times across the video? - check pacing: does the information density vary (compression, expansion, shock cycles) or is it flat? - check authenticity signals: does the script pass the "would claude write this by default?" test? are there AI slop vocabulary patterns present? - produce a section-by-section diagnosis with specific rewrite recommendations ## output standards - always cite specific data from nexlev tools to support conclusions - use specific numbers, not ranges or approximations - name the exact problem before proposing the fix - rank recommendations by expected retention impact, highest first - never recommend changing niche as the first action. diagnose the script layer first

the workflow in practice opus 5

once the skill file is loaded into a opus 5 project with nexlev MCP connected, you can run any of the four analysis modes with a single prompt.

example 1: full channel audit

"run a full channel audit on [channel name or URL]. i want to understand why our last 10 uploads have averaged 15,000 views when our outlier videos have hit 200,000+. identify the structural differences between the outliers and the underperformers and give me the top 3 changes that would move the average."

opus 5 pulls the channel data through nexlev, runs the comparison across all 30 recent videos, identifies the outlier patterns, cross-references them against the underperformers, and delivers a priority-ranked strategy document.

a human analyst takes 3-5 hours to produce the same deliverable. opus 5 does it in one conversation.

example 2: competitive benchmarking

"benchmark my channel [URL] against these 4 competitors: [URLs]. show me where i'm winning, where i'm losing, and where the blue ocean opportunities are that none of us are covering."

opus 5 pulls data on all 5 channels simultaneously, maps the topic overlap, identifies the gaps, and produces a competitive positioning map. the output tells you exactly which topics to cover next and why, backed by specific data from nexlev.

if you want to go deeper on how the retention diagnosis mode feeds directly into the script rewriting process, i cover that in the scriptwriting newsletter at fyreinteractive.co/newsletter.

example 3: niche opportunity scan

"scan for untapped niche opportunities in the [broader category] space. i want channels with fewer than 30,000 subscribers that are pulling 100K+ average views. filter by RPM above $8 and show me the content format driving the outliers."

this query used to require 2-3 hours of manual filtering across multiple tools. opus 5 with nexlev MCP runs it in minutes because it has direct access to the channel database and can apply multiple filters simultaneously.

why opus 5 specifically (and not sonnet 5 or opus 4.8)

three things make opus 5 the right model for analytics work:

long-horizon reasoning. the channel audit requires holding 30+ videos of data in context while cross-referencing patterns across outliers and underperformers. opus 5's reasoning depth on complex analytical tasks is a 10-point jump over opus 4.8 on hex's core analytics benchmark, the first model to break 90%.

the 1M token context window. you can feed the full analytics dataset from nexlev alongside multiple video transcripts and have opus 5 cross-reference everything in a single pass. no chunking. no losing context between steps.

self-checking output. opus 5 validates its own analysis without being asked. when it identifies a pattern, it checks whether the data actually supports the conclusion before presenting it. this is the difference between a model that generates plausible analysis and one that generates accurate analysis.

the cost is higher ($10/$50 per million tokens vs sonnet 5's $2/$10), but for a task that replaces a $5-10K/mo consultant, the economics are not even close. a full channel audit on opus 5 costs under $5 in tokens. the same deliverable from a human costs $500-$1,000.

building this into a repeatable system

the skill file above is a standalone tool. you can run it as-is and get genuine analyst-level output.

but the real leverage comes when you connect the analytics output to the scriptwriting workflow. the channel audit identifies which structural patterns are working. the competitive benchmark identifies which topics to cover. the retention diagnosis identifies what to fix. and all of that feeds directly into the script brief that produces the next batch of videos.

this is the loop that FacelessOS runs for 300+ creators: research identifies the opportunity, analytics diagnoses the gap, the script brief encodes the fix, and the skill files produce the script. each cycle is informed by the data from the previous one.

ed ran this loop across two faceless channels started from zero in january. $50,000+ in 2 months. not because he found a better niche. because each batch of scripts was informed by the analytics from the previous batch.

joey's 863,000-view script came from an outlier analysis that identified a structural pattern in his niche that no competitor was using. the analytics pointed to the opportunity. the script architecture captured it.

if you want the full system, the analyst skill file plus 15 other skill files plus the 6 power workflows, all calibrated across 8,000+ scripts and 50+ niches, grab FacelessOS at fyreinteractive.co/facelessos

(8,000+ scripts. $5M+ generated for clients. 22 skill files trained on pattern data across 50+ niches.)

haris