The master production standard for every ETKM project — regardless of type, tool, or deliverable. Load this skill at the start of any new project build, before any asset is created, any code is written, or any brief is sent to another tool. This is the packaging layer. It does not replace etkm-brand-kit, etkm-deliverable-qc, or etkm-format-standards — it governs when and how those skills are applied. No project enters production without a completed Project Card. No deliverable leaves production without passing QC. No handoff happens without a completed Brief. These three rules are absolute.
Version: 1.0 Established: 2026-03-23 Authority: Nathan Lundstrom / East Texas Krav Maga Benchmark: P1 Book Intelligence System
Every tool in the ETKM stack — Claude Chat, Claude Code, Manus, Cowork — produces better output when it receives better input. The root cause of inconsistent quality is not a capability problem. It is a packaging problem.
When a project enters production without a locked spec, tools improvise. When a handoff brief is missing, the receiving tool fills the gaps with defaults. When QC criteria are not written before the build, "done" becomes subjective.
This skill is the permanent fix.
It establishes one standard container that every project — regardless of type or tool — gets built inside. The container defines what goes in, what comes out, who does each job, and what "done" looks like before anyone starts building.
The P1 Book Intelligence System is the benchmark. It works at a consistently high level because it has all six layers locked before any session runs. Every project we build from this point forward gets the same treatment.
Rule 1: No project enters production without a completed Project Card. Rule 2: No deliverable leaves production without passing all QC gates. Rule 3: No handoff happens without a completed Handoff Brief. Rule 4: No tool improvises brand, voice, or format decisions. Rule 5: The benchmark for every project is the best version of that project type — not the average.
Every project is built by completing six layers in sequence. Do not begin Layer 2 until Layer 1 is complete. Do not begin production until all six layers are complete.
Define the project completely before touching any tool.
Six questions. Every question answered before moving to Layer 2.
Q1: What does this project receive? The exact format of the input. What comes in — document, data, brief, content draft? What format? What minimum quality standard must the input meet before the project runs? If the input is wrong, the output will be wrong. Define acceptable input here.
Q2: What does this project produce? Every deliverable named. Not "a PDF" — which PDF, what length, what sections, what format. Not "an email sequence" — how many emails, what arc, what platform, what format. Be specific enough that the output could be QC'd against this description without asking the builder any questions.
Q3: What is this project looking for? The extraction or processing framework. What does the agent mine from the input to produce the output? What are the key signals, structures, or data points that drive the deliverables? This is the intelligence layer — what makes the output smart vs. generic.
Q4: What brand and voice rules govern this project? Which skills load. What tone. What audience. What prohibited language. What required phrasing. This layer locks before any copy is written so no tool makes brand decisions independently.
Q5: What are the rejection criteria? Write the failure conditions before the build starts. What does a failed deliverable look like for this project specifically? Add project-specific QC items to the universal gates in etkm-deliverable-qc. If you cannot describe failure, you cannot QC for it.
Q6: What is the north star? The single governing standard every deliverable is measured against. One sentence. Not a goal — a standard. P1's north star: "Learning only happens if there is changed behavior." Every asset in P1 is measured against that one sentence. Every project needs its own.
The completed Project Card is the single-page identity document for the project. It is written once, stored in the project skill, and loaded at the start of every session that touches this project.
PROJECT CARD
─────────────────────────────────────────────────────────
Name: [Project name]
Code: [P# or WF-###]
Type: [PDF / Email / HTML / App / Campaign / System]
Owner: Nathan Lundstrom
Status: [Scoping / In Production / Complete / Live]
North Star: [One sentence — the governing standard]
Purpose: [One sentence — what this project does]
─────────────────────────────────────────────────────────
INPUT
Format: [Exact description of what comes in]
Minimum Std: [What the input must contain to be acceptable]
Provided by: [Nathan / Claude / Manus / External]
─────────────────────────────────────────────────────────
OUTPUT
Deliverables: [Every asset named, one per line]
[Asset 01: Name — format, approx length, destination]
[Asset 02: Name — format, approx length, destination]
Format Rules: [Load etkm-format-standards — specify which formats apply]
Destination: [Where the output goes — Nathan / Pipedrive / WordPress / GitHub]
─────────────────────────────────────────────────────────
SKILLS LOADED
Strategy: [etkm-brand-foundation, etkm-audience-map, etc.]
Voice: [etkm-brand-kit, etkm-messaging-playbook, etc.]
Format: [etkm-format-standards + specific format skill]
QC: [etkm-deliverable-qc — always]
Domain: [Project-specific skills: etkm-book-intelligence, etc.]
─────────────────────────────────────────────────────────
AGENT STACK
[See Layer 3 — Agent Stack]
─────────────────────────────────────────────────────────
QC GATES
Universal: etkm-deliverable-qc — all 9 gates
Project-Spec: [Additional rejection criteria specific to this project]
Benchmark: [The best existing version of this deliverable type]
─────────────────────────────────────────────────────────
NORTH STAR
[The single governing standard. One sentence.]
─────────────────────────────────────────────────────────
Define which agent handles which job and what they receive and return.
The agent stack is the internal team for this project. For Chat-only projects it describes the roles Claude plays sequentially. For multi-tool projects it defines which tool owns which phase and what travels between them.
Every project has at minimum four agents:
| Agent | Role | Receives | Returns |
|---|---|---|---|
| Strategist | Defines structure, approach, audience alignment | Project Card | Outline + strategic direction |
| Writer/Builder | Produces the content or build | Outline + brand rules | Draft deliverables |
| Brand Editor | Applies voice, visual, format standards | Draft deliverables | Branded deliverables |
| QC Agent | Holds against all gates — rejects or passes | Branded deliverables | PASS or FAIL with findings |
Complex projects add specialized agents:
| Agent | Role | When to Add |
|---|---|---|
| Architect | Structures multi-asset projects | 3+ deliverables |
| Extractor | Mines raw input for key data | Book, research, or document processing |
| Code Agent | Builds scripts, apps, APIs | Any Claude Code deliverable |
| Deployment Agent | Loads to Pipedrive, WordPress, GitHub | Any Manus deliverable |
Agent Stack Rules:
Gates are checkpoints between phases. Nothing moves through a gate without a pass.
Universal Gate Structure:
GATE A — Input Validation
Before any work begins.
Confirm input meets the minimum standard defined in the Project Card.
If input is insufficient → return to Nathan with specific list of what's missing.
Pass criteria: Input matches the format and minimum standard defined in Layer 1 Q1.
GATE B — Structure Review
After Architect/Strategist completes outline.
Confirm structure matches output spec and north star.
Before Writer/Builder begins.
Pass criteria: Every deliverable in the output spec is accounted for in the structure.
GATE C — Draft Review
After Writer/Builder completes draft deliverables.
Confirm content is complete, on-voice, and on-spec.
Before Brand Editor begins.
Pass criteria: All content present, no placeholder tokens, voice review passes.
GATE D — Brand Review
After Brand Editor completes branded deliverables.
Confirm visual standards, format standards, typography, color.
Before QC Agent runs full gate check.
Pass criteria: Matches etkm-format-standards for the deliverable type.
GATE E — QC Full Run
After Brand Editor — mandatory before any delivery.
Run all 9 gates from etkm-deliverable-qc plus project-specific gates.
Any failure = return to the responsible agent, fix, re-run QC.
Pass criteria: All gates PASS with no exceptions.
GATE F — Delivery
After QC full run passes.
Confirm destination, format, file naming.
Present to Nathan or trigger handoff brief.
Pass criteria: Files named correctly, delivered to correct destination, receipt logged.
Gate Failure Protocol:
Any project that crosses a tool boundary requires a Handoff Brief before the receiving tool begins work. No tool starts blind.
Load etkm-ai-roles before writing any Handoff Brief.
Use when: the project requires a script, API, app, or server-side build.
CLAUDE CODE BRIEF
─────────────────────────────────────────────────────────
Task: [PROJECT-WF-###] [task name]
Status: APPROVED — ready to build
WHAT TO BUILD
[Plain-language description of what the script/app/API does]
INPUT
Format: [Exact data format coming in]
Source: [Where it comes from]
Sample: [Attach sample or describe precisely]
OUTPUT
Deliverable: [Exact file, endpoint, or return value]
Format: [File type, data structure, naming convention]
Destination: [Where it goes when done]
TECHNICAL SPEC
Language: [Python / Node.js / other]
Libraries: [Specific libraries required]
Endpoints: [If API — exact endpoint names and parameters]
File path: [Where in the repo]
ACCEPTANCE CRITERIA
[Numbered list — specific, testable conditions that confirm it works]
1.
2.
3.
QC REQUIREMENT
Run etkm-deliverable-qc on any document output.
For apps: run against acceptance criteria list above.
RECEIPT
Write [PROJECT-WF-###]-COMPLETE.md to Google Drive /ETKM-AI/Status/
Include: what was built, file location, any deviations from spec.
─────────────────────────────────────────────────────────
Use when: the project requires browser automation, Pipedrive setup, WordPress deployment, Make.com build, or any platform-level action.
MANUS BRIEF
─────────────────────────────────────────────────────────
Task: [PROJECT-WF-###] [task name]
Status: APPROVED — ready to deploy
Copy locked: YES — do not modify any copy
ASSETS ATTACHED
[List every file being handed to Manus with its name and location]
- [filename] → [Google Drive path or GitHub path]
STEPS
[Numbered, exact sequence. No ambiguity. One action per step.]
1.
2.
3.
DEPENDENCIES TO VERIFY BEFORE STARTING
[List every URL, field, credential, or condition that must exist first]
-
-
SUCCESS CRITERIA
[What done looks like — exactly. Manus confirms each item.]
-
-
IF SOMETHING DOES NOT MATCH THE SPEC
Stop. Do not improvise. Return to Claude with specific question.
RECEIPT
Write [PROJECT-WF-###]-COMPLETE.md to Google Drive /ETKM-AI/Status/
Include: what was deployed, live URL if applicable, any issues encountered.
─────────────────────────────────────────────────────────
Use when: the project requires background monitoring, file routing, or ongoing watch tasks.
COWORK BRIEF
─────────────────────────────────────────────────────────
Task: [PROJECT-WF-###] [monitoring task]
WATCH
Folder/file: [Exact Google Drive path]
Watch for: [What Cowork is looking for — receipt file, update, stall]
Interval: [How often to check]
TRIGGER CONDITION
[Exactly what Cowork sees that triggers an action]
ACTION ON TRIGGER
[Exactly what Cowork does — alert Nathan, route file, update status]
ALERT FORMAT
Notify Nathan: [What the alert says]
RECEIPT
Update [PROJECT-WF-###] status in registry when trigger fires.
─────────────────────────────────────────────────────────
Before the first asset is built, name the benchmark.
The benchmark is the best existing version of this deliverable type in the ETKM system. If no existing benchmark exists, the benchmark is the P1 Book Intelligence output standard applied to this project type.
Benchmark Questions:
The benchmark is named in the Project Card and referenced by the QC Agent during Gate E. "Does this match or exceed the benchmark?" is the final standard before delivery.
As projects are completed through this framework, their Project Cards become reusable templates for the next project of the same type. This section grows over time.
P1 — Book Intelligence (BENCHMARK)
PDF Production (IN DEVELOPMENT)
Additional project types are added here as their Project Cards are completed.
Claude Chat: Load this skill at the start of any session involving a new or in-progress project. Confirm Project Card exists. If it doesn't — build it before doing anything else.
Claude Code: Receive Handoff Brief. Confirm all acceptance criteria are understood before writing a single line. If anything in the brief is ambiguous — ask before building.
Manus: Receive Handoff Brief. Confirm all steps are clear and all assets are present before executing. Never modify copy. Never improvise steps not in the brief.
Cowork: Receive Monitoring Brief. Watch for triggers. Alert Nathan. Log receipts.
This skill governs when and how all other production skills are loaded.
| Layer | Skills That Load |
|---|---|
| Project Identity | etkm-audience-map, etkm-brand-foundation, etkm-problem-solution |
| Brand & Voice | etkm-brand-kit, etkm-brand-foundation, etkm-messaging-playbook |
| Output Format | etkm-format-standards + format-specific skill (pdf, docx, html, email, pptx) |
| Agent Stack | etkm-ai-roles (for multi-tool projects) |
| QC | etkm-deliverable-qc (always — no exceptions) |
| Handoff | etkm-ai-roles, etkm-pipedrive-manus (for Manus handoffs) |
| Domain | Project-specific skill (etkm-book-intelligence, etkm-lead-gen, etc.) |
Version 1.1 — Updated 2026-03-29
Before creating any new skill, answer these three questions:
Only if the answer to #3 is yes does a new skill get created. All new skills must follow the 6-section template defined in etkm-system-governance.
Frequency: Once every 3 months (first review: June 2026) Duration: 30 minutes
Full protocol defined in etkm-system-governance Section 7. Covers skill inventory, version audit, Notion database health, laptop sync, and system metrics.