+members · pasteable artifacts updated may '26

The snippets library.

Voice files. Prompts. Code. Schemas. Configs. The pasteable artifacts the 12 how-tos point at.

Every file here is the version we ship in our own work — not a generic template, not a tutorial. Paste a voice file into a Claude Project. Drop a schema into your codebase. Wire a handler into your helpdesk. The guides explain the why; these are the what you actually paste.

№ 01

Voice files

3 files

txt voice file

The newsletter voice file.

Three rotating drafting prompts, banned phrases, anchoring samples. The full template behind a one-person newsletter that doesn't sound like AI.

→ Guide № 001 ↓

txt voice file

The cold email voice file.

5-line skeleton, the 90-second reason rule, banned phrases, three real anchoring emails. Different shape than the newsletter voice — same discipline.

→ Guide № 010 ↓

txt voice file

The universal voice file.

Adapt-it-yourself version. Drop into any Claude / GPT / Gemini project. Update monthly — voice drifts.

→ all guides ↓

№ 02

Prompts

6 files

md prompt

Newsletter drafting prompts.

The three rotating prompts that put pressure on the writing — earn-their-place, unbacked claims, adjective swap. Plus the Tuesday cut prompt.

→ Guide № 001 ↓

md prompt

Cold email batch drafting.

The 40-email batch prompt. Plus trigger sources, reply-rate benchmarks, and the manual edit pass that keeps it from smelling AI.

→ Guide № 010 ↓

md prompt

Inbox triage prompts.

The 9am sweep, the reply-in-my-voice prompt, the auto-archive regex list, the end-of-day close ritual.

→ Guide № 004 ↓

md prompt

Research agent system prompt.

The full system prompt for the overnight agent — ground rules, output shape, kill-switch protocol, cost discipline.

→ Guide № 002 ↓

md prompt

Support classifier prompt.

Full system prompt, output schema, labeling rules, and monitoring guidance. Ship one in an afternoon.

→ Guide № 009 ↓

md prompt

Landing page rebuild prompts.

7-section skeleton. Hero variants generator. The cut-30 prompt. The quarterly rebuild diagnostic.

→ Guide № 006 ↓

№ 03

Code & schemas

5 files

ts code

Inngest durable step.

Full pattern for long-running agents — step caching, kill-switch checkpoints, cost watcher, non-retriable errors.

→ Guides № 002, 008 ↓

ts code

Langfuse tracing pattern.

One trace per agent run, one span per step. The required metadata tags so cost dashboards group properly.

→ Guides № 002, 011 ↓

ts code

Classifier webhook handler.

Full handler — idempotency, retry on bad JSON, override rules that beat the model, action layer for any helpdesk.

→ Guide № 009 ↓

sql schema

Cost tagging schema.

Postgres schema, indexes, cost function, and the three dashboards (per-feature, per-user, per-prompt-version). With WoW anomaly query.

→ Guide № 011 ↓

json schema

Classifier output schema.

JSON schema + working examples for routine, urgent-angry, and low-confidence tickets. Plus the override rules that beat the model.

→ Guide № 009 ↓

№ 04

Configs & settings

2 files

md config

Veo prompt scaffold.

Six-axis hero scaffold, motion scaffold, transition scaffold. ElevenLabs locked voice config. Topaz Video AI export settings.

→ Guide № 005 ↓

md config

Descript caption overrides.

Five overrides that kill the default Descript look. Per-platform position presets. Save once, reuse forever.

→ Guide № 007 ↓

№ 05

Checklists

2 files

md checklist

Self-service stack audit.

90-minute version of the paid audit. Inventory, workflow review, risk audit, vendor calendar, ROI math. Surfaces ~80% of what we'd find.

→ paid audit ↓

md checklist

SEO audit for 2026.

Citation tracker template, what still works, what stopped working, sample strategy for 500-keyword libraries, quarterly kill list.

→ Guide № 012 ↓