v2.6 "Agent Infrastructure" shipped 2026-04-21

Stop re-explaining yourself to Claude.

BriefDay turns Claude Code into a daily operating system. Get briefed every morning in 5 minutes. Never lose a decision, contact, or idea again. Now with memory guardrails, EVAL-terminated skills, and a three-category agent taxonomy lens. One ZIP. 15 minutes to set up.

$67 one-time. Launch pricing. No subscription.

Built by a music tech executive who runs his entire business through this system.

claude
$ claude
> /start

Good morning. Here's your briefing:

# Memory loaded (12 active items)
# Inbox: 3 new items (1 flagged high-priority)
# RSS: 47 articles scored, 6 must-reads
# Calendar: 2 meetings today
# Developing story: competitor launched new feature (day 4)

Today's focus: Ship the pricing page.
Wisdom: "Act now. I will act now." (Scroll IX)
20
AI Skills
5
Daily Commands
10
Python Scripts
3
of 6 Agent Layers

Six wins across Memory, Tool Access, and Orchestration.

All dogfooded in a production business for weeks before shipping. No vapor, no AI-generated "ideal workflow" theater.

🛡️

Primacy/Recency Anchors

Every skill file opens and closes with a MANDATORY RULES block. Long-context models ignore middle-of-file rules, so boundary anchoring reclaims compliance.

Layer: Tool Access

EVAL Terminators

Every skill output ends with EVAL: PASS or EVAL: FAIL [reason]. A Stop hook flags missing evals. Warning in v2.6, hard-fail in v2.7.

Layer: Orchestration

📌

Memory Guardrail Hook

SessionStart hook warns when memory.md crosses 180 lines, hard-warns at 200. Prevents silent context bloat as your project count grows.

Layer: Memory

🧠

Agent Taxonomy Lens

Three-category decision lens (Cat1 workflow, Cat2 ReAct, Cat3 multi-agent). Default answer: "No, it's Cat1." Kills 60-70% of over-engineered agent proposals.

Layer: Orchestration

📊

Reliability Calculator v1.1

Now self-classifies each component into Cat1/2/3 with a confidence score. Computes compounded uptime per category. You can override any classification.

Layer: Orchestration

🔎

/lint-brain Skill

Nine memory-integrity checks: broken pointers, orphan files, stale projects, inbox rot, duplicate entries, line budget, cross-ref overlap. Reports only, never auto-fixes.

Layer: Memory

Existing buyers: your download link in the original confirmation email always serves the latest ZIP. Re-download to get v2.6.

Not another chatbot wrapper.

BriefDay is a complete configuration system for Claude Code. Skills, commands, agents, memory, RSS intelligence, and a daily rhythm that actually sticks.

5 Daily Commands

/start, /sync, /wrap-up, /daily, /month-end. Your entire day in three touchpoints taking under 10 minutes total.

🧠

Persistent Memory

Claude remembers your decisions, contacts, active projects, and intel across sessions. No more re-explaining context.

📨

Inbox Processing

Drop files, articles, screenshots, PDFs into inbox/. BriefDay categorizes, scores, extracts insights, and archives everything.

📡

RSS Intelligence

14 starter feeds with scoring, keyword boosts, and deep-read delegation. Surfaces only what matters to your business.

👥

Multi-Agent System

Content reviewer catches AI writing tells. Intel analyst deep-reads articles in background. Model routing keeps costs down.

🔒

Safe Mode + Agent Safety

Hooks block destructive commands. Blast radius discipline on multi-file changes. Turn limits prevent agent runaway. Security audit skill checks your products before launch.

✍️

Voice Calibration

Anti-AI writing rules, ban lists for corporate fluff, position calibration so Claude writes at your level (not a beginner's).

📚

Wisdom System

Spaced-repetition motivation from 8 classic books. Context-aware triggers surface the right wisdom at the right moment.

🔧

20 Skills

LinkedIn ghostwriter, research digest, pitch deck builder, opportunity scanner, quality gate, skeptic review, reliability calculator, lint-brain, and more. All auto-activate.

🌐

Chrome Extension

Right-click any link or page to save it to your inbox. Next /sync picks it up, scores it, and archives it. Two-click capture from anywhere.

📁

Desktop Drop Zone

Create an inbox shortcut on your Desktop. Drag articles, PDFs, screenshots straight in. BriefDay processes everything at /start and /sync.

15 minutes. That's it.

Unzip, run the setup script, fill in your context. Claude Code handles the rest.

1

Unzip and Run Setup

The setup script asks your name, role, industry, and expertise level. Creates all directories and personalizes your config.

./setup.sh
2

Fill In Your Context

Open CONTEXT.md and replace placeholders with your real info. The more specific you are, the better the AI performs.

CONTEXT.md
3

Type /start

Open Claude Code and type /start. Your first personalized briefing runs in under 5 minutes. Drop a link in inbox/, type /sync, and watch it process. Memory builds over 5-7 sessions.

/start

20 skills that auto-activate.

Each skill is a markdown file with YAML frontmatter. Claude loads the right one based on what you're doing. No manual switching.

Skill What It Does
inbox-processorScores, categorizes, and archives inbox items with AI triage
linkedin-ghostwriterDrafts LinkedIn posts in your voice with hook optimization
research-digestSynthesizes articles and research into actionable briefs
rss-readerFetches and scores RSS feeds with keyword relevance
opportunity-scannerScans intel for business opportunities and partnership signals
context-engineeringRORE framework for building high-quality AI context
quality-gateMulti-pass quality checks before any deliverable is final
skeptic-reviewAdversarial review that stress-tests ideas and proposals
content-repurposerTurns one piece of content into multiple formats
product-auditorAudit any product for positioning, pricing, and market fit
planner-worker-judgeMulti-agent pattern: plan, execute, review
product-forgeFull product ideation to build pipeline
pitch-deck-builder3-phase pitch deck pipeline (Scout/Architect/Builder)
outreach-forgeB2B prospecting with parallel research agents
presentation-systemHTML slide decks with branded themes
security-audit3-category security, resilience, and scale audit with fix prompts
reliability-calculatorMap agent stack to six infrastructure layers, score weakest link
lint-brain9 integrity checks on your memory system (broken pointers, orphans, rot)
open-loop-auditSurface stalled projects and unresolved decisions
knowledge-searchFull-text search across your entire knowledge base

Before you build an agent, ask this.

60-70% of "I need an agent" problems are actually Category 1 deterministic workflows. Misclassifying costs you 10x in infrastructure, tokens, and debugging time. Here's the three-category lens BriefDay uses internally.

1

Cat1: Deterministic Workflow

Fixed sequence, same input same output. No branching reasoning. Ships as a skill, hook, or bash script. Default answer: yes, it's Cat1. Typical uptime ceiling: 99.9%+.

Examples: inbox scoring, RSS fetch, daily brief generation, em dash detection, lint checks.

2

Cat2: ReAct Agent

Tool loops with judgment. Iterative refinement, unpredictable paths, LLM decides next step. Ships as a subagent. Only promote here when Cat1 cannot model the problem. Typical uptime: 95-99%.

Examples: deep research, competitive analysis, adversarial review, exploratory web scraping.

3

Cat3: Multi-Agent Network

Specialized agents collaborating. Red flag if proposed. Compounds handoff risk and failure modes. Almost never the right answer. Typical uptime: 90-98%.

Examples: truly open-ended goals with mandatory specialization. If you are not sure, it is not Cat3.

The rule

Default answer to "should this be an agent?" is "No, it's a Category 1 workflow." Only promote to Cat2 when the workflow literally cannot handle the branching. Never jump to Cat3.

Common failure mode: benchmarking Cat1 tools against Cat2 agents on the same spreadsheet. They are architecturally different systems and should not be compared on uptime or cost. BriefDay's reliability-calculator skill runs this classification on your stack and computes compounded uptime per category.

Three kinds of builders.

BriefDay works for anyone who uses Claude Code and wants more than a chatbot.

The Builder

Power user, dev, founder
"I already use Claude Code daily. I want a system, not a blank terminal. BriefDay gave me memory, agents, and a daily rhythm in 15 minutes."

The Operator

Exec, consultant, knowledge worker
"I don't code. I drop things in a folder, type /start, and get briefed. The inbox processing alone saved me an hour a day."

The Explorer

New to Claude Code, curious
"I had Claude Pro but didn't know what to do with Claude Code. BriefDay showed me. Now I run /sync between meetings."

BriefDay is probably not for you if:

  • ✕  You don't use Claude Code (or don't plan to)
  • ✕  You want a drag-and-drop GUI (this is terminal + markdown)
  • ✕  You need real-time team collaboration (this is a single-user system)
  • ✕  You're looking for a ChatGPT wrapper or basic chatbot

This isn't a weekend project.

BriefDay was built by a music tech executive who runs multiple companies through this system daily. 20+ years in digital distribution, platform ops, and revenue strategy. Former SVP at a major Latin label. Co-founder of a boutique distribution company.

Every skill, command, and workflow in BriefDay exists because it solved a real problem in a real business. The memory system, the RSS intelligence, the inbox processing, the content pipeline: all battle-tested across 12+ months of daily sessions.

You're not buying a template. You're buying a year of refinement.

One price. Everything included.

Less than one month of most AI subscriptions. Pays for itself in the first week. No tiers. No recurring fees.

Full System
$67
One-time payment. Yours forever.
  • CLAUDE.md + CONTEXT.md templates (personalized by setup script)
  • 5 daily commands (/start, /sync, /wrap-up, /daily, /month-end)
  • 18 auto-activating skills
  • 2 custom agents (content-reviewer, intel-analyst)
  • 12 Python scripts (RSS, inbox, entity scanning, knowledge search, sync state)
  • Persistent memory system + templates
  • Voice calibration + anti-AI writing rules
  • Wisdom system (8 books, spaced repetition)
  • 14 RSS starter feeds + scoring engine
  • 3 industry examples (consultant, developer, founder)
  • Safe mode hooks (destructive command blocking)
  • Cross-platform setup script (macOS, Linux, WSL)
  • Chrome extension (right-click any link to inbox)
  • Desktop inbox shortcut setup
Get BriefDay - $67

14-day money-back guarantee. No questions asked.

By purchasing, you agree to our Terms of Service.

Not sure yet? Start with the Starter Kit.

Free

Starter Kit

The foundation BriefDay is built on. See how a real CLAUDE.md works before committing.

  •   CLAUDE.md template
  •   Voice guide with anti-AI rules
  •   /start command
  •   Memory template
  • ✕  No skills, agents, or scripts
  • ✕  No RSS, inbox, or intelligence
  • ✕  No daily workflow commands

One email with the download. No spam.

$67 one-time

Full System

Everything in the Starter Kit, plus the complete operating system.

  •   Everything in Starter Kit
  •   18 auto-activating skills
  •   2 custom agents
  •   12 Python scripts
  •   RSS intelligence + inbox processing
  •   5 daily commands + full workflow
  •   Chrome extension + desktop drop zone
Get BriefDay - $67

14-day money-back guarantee.

Changelog

v2.6 "Agent Infrastructure" - April 2026

  • Primacy/recency anchors - Every skill opens and closes with a MANDATORY RULES block. Long-context models ignore middle-of-file rules; boundary anchoring reclaims compliance.
  • EVAL terminators + Stop hook - Every skill output ends with EVAL: PASS or EVAL: FAIL [reason]. Warning mode in v2.6, hard-fail in v2.7.
  • Agent taxonomy decision lens - Three categories (Cat1 workflow, Cat2 ReAct, Cat3 multi-agent). Default: "No, it's Cat1." Kills over-engineering.
  • Memory guardrail hook - SessionStart warns at 180 lines, hard-warns at 200. Prevents silent context bloat.
  • Reliability Calculator v1.1 - Self-classifies your components into Cat1/2/3 with confidence scores. Overridable.
  • /lint-brain skill - Nine memory-integrity checks (broken pointers, orphans, stale projects, inbox rot, duplicates, line budget, cross-ref overlap, stale topics). Reports only.
  • 20 skills, CLAUDE.md v2.6, 3 of 6 Agent Infrastructure layers covered.

v2.5 - April 2026

  • Token discipline - Model routing table (Haiku/Sonnet/Opus by task type), fresh-conversation guidance, explicit model routing for subagents. Cut token waste 40-60%.
  • Session briefs - /wrap-up now generates a session-brief.md (under 2KB) capturing active items + corrections. /start fast-loads it, skipping 4+ file reads on boot.
  • Skill lifecycle - New _archived/ directory for inactive skills. /month-end skill audit moves unused skills out. Fewer active skills = better auto-activation.
  • Lessons compression - /month-end now compresses lessons.md when it exceeds 100 lines. Archives full version, keeps only high-recurrence patterns.
  • New rules - Agent failure logging (fix root cause, don't retry blindly), project context loading (read project memory before any project task).
  • 18 skills, CLAUDE.md v2.5

v2.3 - March 2026

  • Self-learning RSS scoring - Auto-prunes low-signal patterns, compiled pattern cache for 50% faster scoring, monthly weight decay prevents stale keywords.
  • Open Loop Audit skill - Classifies backlog items as "leaves desk" (agent completes) vs "lands on desk" (you decide). Catches delegation theater.
  • WebFetch blocklist - 4-tier domain blocklist prevents wasted fetches on paywalls and bot-blocked sites. Pre-loaded with common blockers.
  • First 15 Minutes onboarding - Restructured setup guide. See your first briefing in under 5 minutes, customize on Day 2+.
  • Sync-state timing - Per-component performance tracking (RSS, YouTube, entity scan). Detect regressions across sessions.
  • 18 skills, scoring weights template with proper customization fields, CLAUDE.md v2.3

v2.2 - March 2026

  • Security audit skill - 3-category audit (security, resilience, scale) with agent-ready fix prompts. Run before launches or quarterly.
  • Agent safety rules - Blast radius discipline, maxTurns caps, co-evolution audits. Prevents runaway agents and multi-file disasters.
  • /month-end upgrades - Scar tissue audit (prune rules Claude follows by default), outcome vs output check, quarterly security reviews.
  • 18 skills (was 16), CLAUDE.md v2.2

v2.1 - March 2026

  • Full-text knowledge search - FTS5 index across all your documents. Find anything in under a second with BM25 ranking.
  • Intel auto-archiving - intel-log.md stays a clean rolling 30-day window. Older entries archive automatically by month.
  • Entity discovery - entity scanner now runs per-sync with --quick mode. Auto-discovers new entities not on your watchlist.
  • Knowledge search skill - new skill wraps search tools for agent access. "What do I know about X?" now works.
  • 16 skills (was 15), 12 scripts (was 10)

v1.0 - February 2026

  • Initial release: 15 skills, 10 scripts, 5 commands
  • Persistent memory, RSS intelligence, entity tracking
  • Content reviewer and intel analyst subagents
  • Wisdom framework with 8 classic books
  • Chrome extension for inbox capture

Common questions.

Do I need to know how to code?

No. You unzip a folder, run a setup script, and type /start. The config files are markdown (plain text). If you can edit a text file, you can use BriefDay.

What do I need to run this?

Claude Code (requires a Claude Pro, Team, or Enterprise subscription from Anthropic). macOS, Linux, or WSL. Python 3.8+ for the scripts.

Is this safe? Can it delete my files?

BriefDay ships in safe mode. Hooks block destructive commands (rm -rf, force push, etc.). Memory gets backed up before edits. No skip-permissions required.

How is this different from a ChatGPT custom instruction?

Custom instructions are a paragraph of text. BriefDay is an entire system: 20 skills that auto-activate, persistent memory across sessions, RSS feeds, inbox processing, multi-agent delegation, and a daily workflow. It turns Claude Code into an operating system.

What is "Agent Infrastructure" and why does it matter?

Every AI agent depends on six infrastructure layers: compute, identity, memory, tool access, billing, and orchestration. You handle compute and billing (Claude Code + Anthropic subscription). BriefDay solves the three that actually break: memory that persists, tools that work without configuration, and orchestration that doesn't lose your work when context compacts. v2.6 moves eight specific rules from "Claude should do X" to "the system ensures X."

I already bought v2.5. Do I need to pay again?

No. Updates are included in your one-time purchase. The download link in your original confirmation email always serves the latest ZIP. Re-download to get v2.6. Your existing CLAUDE.md will not be overwritten, the setup script will ask before touching any file.

Can I customize it for my industry?

That's the entire point. The setup script personalizes your config. CONTEXT.md is a fill-in-the-blank profile. Add your own RSS feeds, contacts, decision filters, and voice rules. Three industry examples are included.

What if I already have a CLAUDE.md?

BriefDay's CLAUDE.md is modular. You can merge sections into your existing config or use BriefDay's as a fresh start. The setup script won't overwrite anything without asking.

Do I get updates?

Yes. v2.6 "Agent Infrastructure" shipped 2026-04-21 with memory guardrails, EVAL terminators, primacy/recency skill anchors, agent taxonomy lens, and /lint-brain memory integrity checks. Updates are included for all future versions. The system is also designed to be extended by you over time.

What's the refund policy?

14 days, no questions asked. If BriefDay isn't for you, email us and you get a full refund.

Will I actually use this after the first week?

Most people hit the habit around session 5. That's when /sync saves you real time and memory starts carrying context you didn't have to re-type. If you don't see value in 14 days, full refund.

Is $67 really one-time? No hidden fees?

One-time, forever yours. No subscription. You get all updates to the current major version. v2.6 is the latest.

Why not just build my own CLAUDE.md?

You could. It took 12 months to build this one. BriefDay gives you the system, the skills, the scripts, the memory architecture, and the daily workflow out of the box. You customize it. You don't have to build it from scratch.

My role is pretty unique. Will this work for me?

BriefDay is a framework, not a rigid template. The setup script asks your name, role, and industry. CONTEXT.md is fill-in-the-blank. Voice.md adapts to your writing style. Three examples ship (consultant, developer, founder), but the system works for any knowledge worker.

Can my whole team use this?

BriefDay is a single-user system (one license = one person). Each person gets their own config, memory, and context. Looking for a team setup? Email hello@briefday.ai and we'll work something out.