⚠️ Simulation mode active · No live agent runs. Log in to enable live runs against hooks-os.
A
Hooks-OS · AgentsAdmin · Job-First SaaS · v1.0
7 agents live 42/50 eval passing 12 sims today consistency 96%

The job. not the tech.

Every agent is a paycheck workflow — a real job in hooks-os that used to eat Kyron's or Hooks' hours. Not "AI-powered anything." Here is the job your team no longer does by hand. Aim: 9/10 quality with 10/10 consistency, on vanilla intellect, tested against 50 real cases per agent before it ships.

Job Matrix · Top workflows ranked

7 jobs. every one has a paycheck.

Every job below is scored on 5 dimensions per the rubric: how easy to know done · how often · how expensive the pain · what tools · who owns budget. Sorted by pain × frequency. This IS the roadmap.

Job (workflow) Archetype Freq Pain / week Finish line Tools Budget owner Confidence Actions
Inbound Lead Triage
Every DM/email classified + routed
Triage 5-15/day ~3.5 hrs
~$525
Correctly tagged in <5 min IG · Email · data/ Hooks (ops) 9.2/10 · 98%
Athlete Outreach Draft
DMs to college coaches in Hooks' voice
Draft 10-30/wk (peaks) ~4 hrs
~$800
Ships w/ ≤1 line edit 90% of time voice-rules · film · templates Hooks (recruiting) 9.4/10 · 97%
Weekly Monday Brief
Sunday-night draft for Monday 7am
Draft 1/wk ~45 min
~$150
Approved by Hooks in ≤5 min data/ · goals/ · output/ Hooks (ops) 9.6/10 · 100%
Advocacy Call Logger
Voice memo → CBH ledger entry
Coordinate 2-5/wk ~1.5 hrs
~$300
100% calls logged within 2 hrs Voice transcribe · CBH ledger Hooks (attribution) 9.5/10 · 99%
NIL Deal Progression
Stalled deals get moved every Friday
Coordinate weekly per deal ~2 hrs
~$400
Every stalled deal actioned by Fri EOD deals/ · email · comms log Athlete + Hooks (rev) 8.7/10 · 93%
Trainer Blueprint Follow-up
Prospect touches per SLA
Act 5-15 leads live ~2.5 hrs
~$500
100% leads touched within SLA Email send · CRM · Stripe Hooks (TB revenue) 8.4/10 · 91%
Engagement Monitor
Elevated-mode alerts within 30 min
Triage continuous ~30 min
~$100 opp cost
Alerts fire within 30 min of spike IG API · YouTube API Hooks (brand) 8.2/10 · 89%

◆ Pain totals: ~14 hours/week of low-value coordination Hooks/Kyron do by hand. At $200/hr blended = $2,800/week reclaimed capacity when fully deployed.

Live Agents · Full spec per agent

The 7-part spec. every agent.

Trigger · Context · Tools · Rules · Approval · Escalation · Success. If any part is missing, the agent doesn't ship. Trusted agent knows the job, knows limits, knows when to ask.

Triage · Classifies + routes Confidence 9.2/10 · Consistency 98% · 47/50 eval
Inbound Lead Triage

"Here is the classification + routing your inbox no longer does by hand." Every IG DM and email gets a class, a priority, and a route within 5 minutes.

◆ TriggerNew IG DM to @stlfootball or @coachhooks · new email to info@skysthelimit.com · anything with "class of 20X" or "my son plays" pattern
◆ Contextdata/athletes/ (roster + alumni) · data/inbound-emails/ (recent history) · data/deals/ (NIL contacts) · voice-rules-v2
◆ Toolsread: IG DM · Gmail · repo data · write: inbound-emails/{date}-{source}-{sender}.md
◆ Allowed to doClassify (athlete / trainer-TB / NIL / media / other) · tag priority (P0-P3) · draft first-touch reply (draft only) · commit to data/
◆ Success95%+ classification accuracy · 100% touched within 4 hrs · P0 items alerted within 15 min
◆ ApprovalFirst-touch reply drafts require Hooks approve. Classification + routing runs autonomously with weekly Hooks audit.
◆ EscalateConfidence <70% on classification · P0 pattern detected · legal/media inquiry keywords · immediately ping Hooks
Eval Gym: 47/50 passing · updated 2h ago
Draft · Human approves Confidence 9.4/10 · Consistency 97% · 48/50 eval
Athlete Outreach Draft

"Here are the college coach DMs your evenings no longer draft." Hooks says the athlete's name · agent drafts in his voice · Hooks approves + sends. 8-12 min → 30 sec per DM.

◆ TriggerHooks message pattern: "draft outreach for {athlete-slug}" · or scheduled batch for a class
◆ Contextdata/athletes/{athlete}/profile.md · film.md · recruiting.md · voice-rules-v2 · output/outreach/ (last 30 examples)
◆ Toolsread: athlete data · voice rules · HUDL timestamps · write: output/outreach/{date}-{athlete}-{coach}.md
◆ Allowed to doDraft DM in Hooks' voice · include position stack + HUDL timestamp + #DA3WAY sign · commit to output/
◆ Success90%+ drafts ship with ≤1 line edit · never uses banned phrases ("I hope this finds you well" etc.)
◆ ApprovalALWAYS. Every outreach DM requires Hooks approve before send. Zero exceptions.
◆ EscalateThin athlete data (<3 film clips) → ask Hooks for one-line context · unfamiliar coach → surface Prospect Edge notes first
Eval Gym: 48/50 passing · updated 4h ago
Draft · Human approves Confidence 9.6/10 · Consistency 100% · 50/50 eval
Weekly Monday Brief

"Here is the Monday briefing your Sunday no longer writes." 45 min → 5 min approve. Every Sunday 9pm CT, agent drafts. Every Monday 7am, brief is in Hooks' inbox.

◆ TriggerSunday 9:00pm CT cron · Hooks weekly override available
◆ ContextFull week's data/ · goals/weekly.md · output/decisions/ · advocacy ledger delta · engagement spikes
◆ Toolsread: entire repo · write: output/reports/weekly-brief-{date}.md · send: email + Slack to Hooks
◆ Allowed to doDraft brief · propose top-3 priorities · flag anything red (missed follow-ups · stalled deals · engagement dips) · queue for approval
◆ SuccessBrief delivered by 7am Monday · Hooks approves in ≤5 min · zero missed weeks
◆ ApprovalHooks approves the top-3 priorities before Monday 8am. Draft brief goes out to team unread if approved. Silent-approve after 4hr window.
◆ EscalateAny funnel with "urgent" tag → highlight in brief and alert separately · unusual pattern (2× baseline signal in any category) → wake Hooks
Eval Gym: 50/50 passing · updated 6h ago
Coordinate · Keeps work moving Confidence 9.5/10 · Consistency 99% · 49/50 eval
Advocacy Call Logger

"Here is the ledger entry your call notes no longer write." Hooks voice memos the coach call · agent transcribes + extracts + drafts CBH ledger entry within 2 hours.

◆ TriggerHooks voice memo sent to advocacy-log@skysthelimit.com · or triggered post-scheduled-call calendar event
◆ Contextdata/athletes/ (athlete named in call) · CBH ledger (prior entries for this coach) · College Directory (coach + program info)
◆ Toolsread: voice memo transcribe (Whisper) · athlete data · write: draft ledger entry with structured data
◆ Allowed to doTranscribe · extract structured fields (date · coach · school · duration · outcome · next-step) · draft entry · queue for Hooks approve
◆ Success100% of memos logged within 2 hours · zero data errors · zero missed calls in the ledger
◆ ApprovalHooks approves entry before commit to ledger. Silent-approve after 24hr window.
◆ EscalateAmbiguous outcome ("neutral" that could go either way) → ask Hooks · legal-relevant coach comment → alert immediately · never assume signing intent
Eval Gym: 49/50 passing · updated 1d ago
Coordinate · Keeps work moving Confidence 8.7/10 · Consistency 93% · 44/50 eval
NIL Deal Progression

"Here is the Friday follow-up your deal spreadsheet no longer chases." Every Friday 3pm, agent reviews the deal pipeline · drafts follow-ups on anything silent 5+ days · commits by EOD.

◆ TriggerFriday 3:00pm CT cron · plus manual "check {deal-slug}" on-demand
◆ Contextdata/deals/ (all deals) · communications.md per deal · athlete + brand contact info
◆ Toolsread: all deal data · draft: follow-up email · write: deals/{deal}/communications.md · DO NOT send: send always requires Hooks approve
◆ Allowed to doIdentify stalled deals (silent 5+ days) · draft polite follow-up · propose next-step · commit draft to comms log · queue for Hooks approve
◆ Success100% of stalled deals actioned by Fri EOD · zero over-eager follow-ups (never <5 days silence)
◆ ApprovalHooks approves each follow-up before send. NIL is legal territory · zero autonomous sends.
◆ EscalateContract-language keywords in comms → Hooks + Kyron · brand asking about athlete's personal life → Hooks · anything IP or exclusivity → Hooks + legal
Eval Gym: 44/50 passing · 6 failure cases under review
Act · Under direct order Confidence 8.4/10 · Consistency 91% · 43/50 eval
Trainer Blueprint Follow-up

"Here is the sales-sequence follow-up your calendar no longer forgets." Every TB lead in-flight gets touched per its SLA. Sequence is pre-approved · agent runs it autonomously within the approved pattern.

◆ TriggerNew TB lead added to data/projects/trainer-blueprint/leads/ · or SLA hit (3-day · 7-day · 14-day windows)
◆ ContextTB sales sequence (pre-approved 3 emails) · lead's status · prior touches · lead source
◆ Toolsread: lead data · sequence templates · send: email via SMTP · write: lead status update in CRM
◆ Allowed to doSend sequence emails autonomously (pre-approved language) · update lead status · schedule next touch · log all outputs
◆ Success100% of TB leads touched within SLA window · zero over-touches (max 3 sequence emails per lead)
◆ ApprovalSequence pre-approved by Hooks. Individual emails send autonomously. Custom replies (non-sequence) always require Hooks.
◆ EscalateLead replies with question outside sequence · pricing negotiation attempt · asks for demo call → Hooks. Kill sequence + wait.
Eval Gym: 43/50 passing · 7 edge cases spec'd
Triage · Classifies + routes Confidence 8.2/10 · Consistency 89% · 41/50 eval
Engagement Monitor

"Here is the audience-signal read your Instagram no longer misses." Elevated-mode moments (2× baseline engagement) get flagged within 30 min so Hooks can capitalize while the room is hot.

◆ TriggerContinuous monitoring of IG + YouTube engagement · 5-min sampling · baseline computed weekly
◆ Contextdata/engagement/ · historical baseline per platform · recent post content · TB campaign windows
◆ Toolsread: IG API · YouTube API · write: alert to Slack · engagement/{date}-elevated.md
◆ Allowed to doClassify engagement level (baseline · elevated · viral) · surface content that triggered elevation · propose action (post follow-up · pin comment · run TB CTA)
◆ SuccessElevated-mode alert fires within 30 min of spike · zero false positives per week · zero missed viral moments
◆ ApprovalAlert only. Any content action (post follow-up · pin · sequence trigger) requires Hooks approve within the alert.
◆ EscalateSustained 3× baseline engagement (viral moment) → phone-call-worthy alert · any crisis-shape comment volume → immediate Hooks + Kyron
Eval Gym: 41/50 passing · baseline tuning in progress
Simulation Harness · Eval Gym

Nothing ships. without 50 real cases.

Every agent has an Eval Gym of 50 real historical examples pulled from hooks-os data. Agent runs against them. Fails get spec'd. Fixed. Re-run. Repeat until 45+/50 pass. Only then does it touch production.

Recent simulation runs · last 24 hours
12 runs · 10 passing · 2 review
14:32 CTMonday Brief · test-week-2026-07-14 — draft matched Hooks' approved v1 within 0.94 simPass
13:18 CTOutreach Draft · test-marvin-mims-vandy — Vandy DM matched approved template + banned-phrase check cleanPass
12:47 CTInbound Triage · test-parent-inquiry-021 — classified correctly · priority P2 accuratePass
12:15 CTNIL Deal · test-bevo-stalled-14d — draft follow-up used wrong first-name (Andrew vs Drew) · spec updatedReview
11:08 CTAdvocacy Log · test-tcu-call-2026-07-12 — outcome classification accurate · next-step correctPass
10:30 CTEngagement Monitor · test-marvin-highlight-post — flagged as elevated within 22 min · action proposal correctPass
09:12 CTTB Follow-up · test-coach-jenkins-day-7 — sequence email fired correctly · attribution loggedPass
08:44 CTOutreach Draft · test-kai-james-oklahoma — draft ok but missed HUDL timestamp inclusion ruleReview
08:20 CTMonday Brief · test-week-2026-07-07 replay — priority ranking matched Hooks' approved orderPass

Data is the moat

Every failure captured in the Eval Gym becomes training data. 50 real cases per agent is the floor. Every miss becomes a rule in feedback/ that prevents it happening again.

Vanilla intellect only

No fine-tuning. No custom models. Whatever Claude ships with — the moat is the data + the spec, not the weights. Portable to any model this way.

9/10 quality · 10/10 consistency

Not aiming for genius. Aiming for never surprise Hooks in a bad way. Predictable is worth more than brilliant.

Sell the pilot labor

Run agent alongside Hooks' current system during the pilot. Show the receipts. "We tested this on your last 50 advocacy calls · here's what we caught." Trust before productize.

Control Room · Customer needs control

The 6 controls. customer trust surface.

Log · Approvals · Settings · Handoffs · Analytics · Policy Editor. What happened. Why. What needs review. Where did it fail. Every question the customer will ask — answered before they ask.

Control 01
Log
2,847
Runs this week: 342
Approvals sent: 89
Actions taken: 213
Escalations: 12
Open log
Control 02
Approvals
8
Monday brief: 1 pending
Outreach drafts: 4 queued
Ledger entries: 2 queued
NIL follow-ups: 1 waiting
Review queue
Control 03
Settings
7
Auto-approve window: 4h (brief) · 24h (log)
Voice rules: v2.3 · 47 rules
Escalation ON: all 7 agents
Sim mode: live agents run parallel
Configure
Control 04
Handoffs
31
Agent → Agent: 24 this week
Agent → Hooks: 12 this week
Agent → Kyron: 3 this week
Legal escalations: 0
See map
Control 05
Analytics
96%
Overall consistency: 96%
Time reclaimed (wk): 14 hrs
$ value (wk): $2,800
Drift alerts: 0
Full dashboard
Control 06
Policy Editor
47
Voice rules: 47 committed
Banned phrases: 12
Auto-approve: per-agent
Escalation rules: 21 active
Edit policies
0-100 Plan · 4 weeks

From spec to sold. 28 days.

The 4-week rollout that takes hooks-os agents from admin dashboard to 3 productized pilots per niche. Every action ordered by pain × frequency × paycheck.

◆ Week 01 · Foundation
Ship the 7 agents
Day 1-2 · Pick paycheck workflows (done · 7 identified)

Day 3-4 · Shadow 20 real runs per agent · screen record · narrate steps

Day 5 · Extract 7-part spec per agent

Day 6 · Build smallest useful version (Draft + Approve loop)

Day 7 · Create Eval Gym · 50 real cases per agent
Now
◆ Week 02 · Pilot
Sell 2 pilots · same niche
Approach 2 elite-tier coaches with the exact same workflow · same pain · same tools · same success metric

Pricing: $1,500 setup + $1,000/mo for 3-month pilot · OR $30/booked-job outcome pricing

Run agent parallel to their current system · show the receipts weekly

Sales asset: "We tested this on your last 50 advocacy calls"
Next
◆ Week 03 · Product
Add the wrapper
Add the product wrapper around the agent set:

· Logs
· Settings
· Approvals queue
· Handoffs
· Analytics
· Policy editor

Design tools: Claude Design + Fable 5

Customer trust surface first · features second
Week 3
◆ Week 04 · Distribute
Content · then ads
Publish workflow tier-down content:

"How Coach Hooks manages 30+ P4 recruits without dropping a call" — long-form video · IG carousel · newsletter

Watch what performs organically

Put paid ads behind the content that already wins organic

Now you know where to scale.
Week 4

Red hot problems only

Every job on this dashboard is a red-hot problem: big budget · existing pain · missed revenue. No lukewarm workflows. If the buyer can't feel the loss, it's the wrong workflow.

Human implementation SaaS

hooks-os requires Hooks in the loop. Every agent has a human moment. That's the moat AI-alone can't cross — pick industries with regulation, credentialing, or relationship trust like this one.

Convert tokens → product at exchange rate

Every hour of Kyron's or Hooks' time = ~$200 blended. Every agent replacement of that time is billed at a fraction. The delta is the margin. Aim: $2K/mo returns $2.8K/wk reclaimed capacity.

Product is the job

Never sell "AI-powered advocacy management." Sell "Your ledger writes itself now." The pitch is the outcome. The product is the job you no longer do by hand.