John Lindquist's workshop, 2026-07-17. This file is the index an agent fetches to navigate the whole workshop: eleven sections, each with a human guide page, a session replay, a Socratic dialogue, and its own machine-executable agent guide.
here.now gotcha (read first): bare .md URLs on here.now serve an HTML viewer page, not markdown. Always fetch markdown files with the ?raw suffix. This file's raw URL is https://ember-bodhi-scrr.here.now/agent.md?raw, and every per-section agent guide below is listed with ?raw already applied.
An agent re-audits last month's field-guide directives against local evidence (conversation history, prototypes, ~2,000 Oracle sessions) to score which laws held and which evolved.
Two files in .pi/ rewire every shell command the agent runs: rm quarantines, git refuses force pushes, linear self-assigns, and every override explains itself on stderr.
One- and two-letter zsh functions wrap flag-heavy agent CLIs and compose into resume, fork, and account-switching families.
An annotated screenshot replaces a paragraph of ambiguous prose: the arrow does the pointing, the words on the image carry the intent.
prompt-builder (p) gives the prompt its own composing surface, then hands off to Pi, Codex, or Claude; handoff mode turns any shell function into a launch target.
A repeatable agent job becomes a markdown file in your repo: context preloaded, ambient config stripped, behavior eval-gated, changes proposal-first.
.md flow files with ?raw)An agentic terminal is a stack: Ghostty draws, Herdr exposes panes over a unix socket, Pi drives the socket, Karabiner makes it one keystroke — up to keeper agents and swarms.
Generate nine wildly different mockups in minutes, pick a direction with your eyes, hand the winner back to an agent as a reference image.
An Elo duel harness turns this-or-that votes into a persistent preference record your agent reads; it is how this workshop's own page theme was chosen.
Oracle bundles context and drives a real logged-in browser into ChatGPT and Gemini: subscription-priced frontier thinking, plus a model that actually watches video.
A press-to-capture recorder pins one sentence of human testimony to a ring buffer of app events and exports a redacted payload any agent can debug from.
@ URL imports) plus all eleven Session Replays and Socratic Dialogues (curl context injections) — then performs one task with all of it in context. Heavyweight by design.mkdir -p flows && curl -sf "https://ember-bodhi-scrr.here.now/flows/workshop.md?raw" -o flows/workshop.md
npx -y mdflow flows/workshop.md --_task "<the user's task>" --engine claude
--engine picks the agent CLI (claude, codex, gemini, pi, …); omit it to use the mdflow default. Verify free first with --_dry-run (no engine call).?raw) and execute it top-to-bottom. Each guide is self-contained, prints every file in full, and ends with a success check — run it and report the result to your user..md URL returns HTML, you forgot ?raw.agent.md (viewer) in their footers; always prefer the ?raw form for machine reads.# Agentic Software Factory — machine-readable agenda John Lindquist's workshop, 2026-07-17. This file is the index an agent fetches to navigate the whole workshop: eleven sections, each with a human guide page, a session replay, a Socratic dialogue, and its own machine-executable agent guide. **here.now gotcha (read first):** bare `.md` URLs on here.now serve an HTML viewer page, not markdown. Always fetch markdown files with the `?raw` suffix. This file's raw URL is https://ember-bodhi-scrr.here.now/agent.md?raw, and every per-section agent guide below is listed with `?raw` already applied. ## Start here: References & Resources - Page: https://dewy-sleet-n5zs.here.now/ - What it is: every tool used in this workshop, with 81 evidence-backed alternatives — use it to map each section onto tools your user already has. ## Sections (in workshop order) ### 00 — How Agentic Work is Evolving An agent re-audits last month's field-guide directives against local evidence (conversation history, prototypes, ~2,000 Oracle sessions) to score which laws held and which evolved. - Guide: https://hallowed-entry-96xv.here.now/ - Replay: https://hallowed-entry-96xv.here.now/replay.html - Dialogue: https://hallowed-entry-96xv.here.now/dialogue.html - Agent guide: https://hallowed-entry-96xv.here.now/agent.md?raw ### 01 — Configuring Your Environment Two files in `.pi/` rewire every shell command the agent runs: `rm` quarantines, `git` refuses force pushes, `linear` self-assigns, and every override explains itself on stderr. - Guide: https://polite-ravine-6v8g.here.now/ - Replay: https://polite-ravine-6v8g.here.now/replay.html - Dialogue: https://polite-ravine-6v8g.here.now/dialogue.html - Agent guide: https://polite-ravine-6v8g.here.now/agent.md?raw ### 02 — Custom zsh Functions One- and two-letter zsh functions wrap flag-heavy agent CLIs and compose into resume, fork, and account-switching families. - Guide: https://sacred-ritual-w7as.here.now/ - Replay: https://sacred-ritual-w7as.here.now/replay.html - Dialogue: https://sacred-ritual-w7as.here.now/dialogue.html - Agent guide: https://sacred-ritual-w7as.here.now/agent.md?raw - Attachments: zsh files — https://sacred-ritual-w7as.here.now/agent-launchers.zsh and https://sacred-ritual-w7as.here.now/pi-utils.zsh ### 03 — Prompting with Images An annotated screenshot replaces a paragraph of ambiguous prose: the arrow does the pointing, the words on the image carry the intent. - Guide: https://amber-willow-n64j.here.now/ - Replay: https://amber-willow-n64j.here.now/replay.html - Dialogue: https://amber-willow-n64j.here.now/dialogue.html - Agent guide: https://amber-willow-n64j.here.now/agent.md?raw ### 04 — Prompt Builder prompt-builder (`p`) gives the prompt its own composing surface, then hands off to Pi, Codex, or Claude; handoff mode turns any shell function into a launch target. - Guide: https://shiny-bloom-f8hv.here.now/ - Replay: https://shiny-bloom-f8hv.here.now/replay.html - Dialogue: https://shiny-bloom-f8hv.here.now/dialogue.html - Agent guide: https://shiny-bloom-f8hv.here.now/agent.md?raw ### 05 — mdflow A repeatable agent job becomes a markdown file in your repo: context preloaded, ambient config stripped, behavior eval-gated, changes proposal-first. - Guide: https://walnut-virtue-6d22.here.now/ - Replay: https://walnut-virtue-6d22.here.now/replay.html - Dialogue: https://walnut-virtue-6d22.here.now/dialogue.html - Agent guide: https://walnut-virtue-6d22.here.now/agent.md?raw - Attachments: flows — listed with raw URLs inside the agent guide (fetch `.md` flow files with `?raw`) ### 06 — Why Pi, Herdr, and Ghostty An agentic terminal is a stack: Ghostty draws, Herdr exposes panes over a unix socket, Pi drives the socket, Karabiner makes it one keystroke — up to keeper agents and swarms. - Guide: https://summit-inlet-kxyf.here.now/ - Replay: https://summit-inlet-kxyf.here.now/replay.html - Dialogue: https://summit-inlet-kxyf.here.now/dialogue.html - Agent guide: https://summit-inlet-kxyf.here.now/agent.md?raw ### 07 — Variations (9x Image Mockups) Generate nine wildly different mockups in minutes, pick a direction with your eyes, hand the winner back to an agent as a reference image. - Guide: https://indigo-kayak-2zw8.here.now/ - Replay: https://indigo-kayak-2zw8.here.now/replay.html - Dialogue: https://indigo-kayak-2zw8.here.now/dialogue.html - Agent guide: https://indigo-kayak-2zw8.here.now/agent.md?raw - Attachments: page skill — under https://indigo-kayak-2zw8.here.now/skills/ (raw URLs inside the agent guide) ### 10 — This or That — RLHF An Elo duel harness turns this-or-that votes into a persistent preference record your agent reads; it is how this workshop's own page theme was chosen. - Guide: https://gentle-lasso-9d5n.here.now/ - Replay: https://gentle-lasso-9d5n.here.now/replay.html - Dialogue: https://gentle-lasso-9d5n.here.now/dialogue.html - Agent guide: https://gentle-lasso-9d5n.here.now/agent.md?raw - Attachments: kit + skill + flow — kit at https://gentle-lasso-9d5n.here.now/kit/, skill and flow raw URLs inside the agent guide ### 08 — Maximum Intelligence and Video Oracle bundles context and drives a real logged-in browser into ChatGPT and Gemini: subscription-priced frontier thinking, plus a model that actually watches video. - Guide: https://shiny-rosette-m6zy.here.now/ - Replay: https://shiny-rosette-m6zy.here.now/replay.html - Dialogue: https://shiny-rosette-m6zy.here.now/dialogue.html - Agent guide: https://shiny-rosette-m6zy.here.now/agent.md?raw - Attachments: Oracle skill family — under https://shiny-rosette-m6zy.here.now/skills/ (raw URLs inside the agent guide) ### 09 — Debugging with Markers A press-to-capture recorder pins one sentence of human testimony to a ring buffer of app events and exports a redacted payload any agent can debug from. - Guide: https://mindful-mango-accy.here.now/ - Replay: https://mindful-mango-accy.here.now/replay.html - Dialogue: https://mindful-mango-accy.here.now/dialogue.html - Agent guide: https://mindful-mango-accy.here.now/agent.md?raw ## The whole workshop as one flow - Flow (raw): https://ember-bodhi-scrr.here.now/flows/workshop.md?raw - What it is: an mdflow flow that preloads this entire workshop as live context — every section's agent guide (`@` URL imports) plus all eleven Session Replays and Socratic Dialogues (`curl` context injections) — then performs one task with all of it in context. Heavyweight by design. - To run it for your user: ``` mkdir -p flows && curl -sf "https://ember-bodhi-scrr.here.now/flows/workshop.md?raw" -o flows/workshop.md npx -y mdflow flows/workshop.md --_task "<the user's task>" --engine claude ``` - `--engine` picks the agent CLI (`claude`, `codex`, `gemini`, `pi`, …); omit it to use the mdflow default. Verify free first with `--_dry-run` (no engine call). ## Notes for agents - To set up any single section for your user: fetch that section's "Agent guide" URL above (already `?raw`) and execute it top-to-bottom. Each guide is self-contained, prints every file in full, and ends with a success check — run it and report the result to your user. - If a fetched `.md` URL returns HTML, you forgot `?raw`. - Section pages link human-facing `agent.md` (viewer) in their footers; always prefer the `?raw` form for machine reads. - This agenda page itself installs nothing; it is an index. There is no success check here beyond confirming the URLs above return HTTP 200.