// 00 — introducing airic

Turn methodology
into executable agents.

Airic is an open-source methodology runtime that converts playbooks, SOPs, expert workflows, and team practices into document-defined agent packs. Agents run the method, reflect on the work, and propose improvements as human-reviewable Markdown diffs.

MIT licensed·Markdown-native·Git-friendly·Workspace-local
~/project/.airic
.airic/
├─ packs/
│  ├─ packsmith/
│  └─ founder-demo/
└─ specs/
   ├─ roles/
   ├─ document-types/
   ├─ processes/
   └─ precedents/
session · founder-demo
01Source·methodology
02Pack·airic extension
03Work·guided execution
04Reflection·review the run
05Patch·spec diff → approve
// 01 — the problem

Agents are powerful. Methods are still trapped in documents.

Modern agents can code, research, browse, and automate. But serious work still depends on methodology: how to frame problems, make decisions, review results, avoid failure modes, and improve over time.

Today, that knowledge is scattered across books, PDFs, Notion pages, consulting decks, SOPs, and expert memory. Skill systems can package procedures, but a real methodology is more than one skill.

books.pdf
notion-wiki
company.sop
deck-v7.key
expert.brain
playbook.md
course-101
team.lore
checklists/

// roles · artifacts · processes · review standards · judgment · learning loops

// 02 — the solution

Airic makes methodology executable.

01

Document-defined agents

Agent behavior lives in editable Markdown, not hidden prompts.

specs/roles/reviewer.md
02

Extension packs

Package a methodology as roles, document types, processes, precedents, and tool policies.

packs/<your-method>/
03

Packsmith

Convert a playbook, SOP, course, or expert workflow into an Airic extension pack.

airic pack from ./source
04

Executable documents

Turn clarified intent into task documents agents can act on.

task.md → run()
05

Visible reflection

Agents review their own work and surface friction, failure patterns, and missing rules.

reflect --run last
06

Human-approved learning

Improvements become reviewable Markdown diffs before changing future agent behavior.

+ added precedent #014
// 03 — packsmith

Packsmith: build packs from methods.

Packsmith is the first Airic extension. It helps you extract a methodology from source material and turn it into a runnable Airic pack.

01 Source Method
02 Method Map
03 Airic Entities
04 Extension Pack
05 Simulation
06 Reflection
07 Improved Pack
example inputs
  • startup playbooks
  • engineering workflows
  • company SOPs
  • consulting frameworks
  • course materials
  • expert interviews
  • existing skill repos
my-method-pack/
my-method-pack/
├─ README.md
├─ roles/
├─ document-types/
├─ processes/
├─ precedents/
├─ tool-policies/
└─ examples/
// 04 — demo

Founder Playbook → Founder Pack

Airic uses Packsmith to convert an AI-native startup playbook into a Founder Demo Pack. This demo proves the core loop: a methodology can be extracted, executed, reviewed, and improved.

  1. 01idea validation
  2. 02problem hypothesis
  3. 03customer discoveryin progress
  4. 04MVP scope
  5. 05launch readiness
  6. 06GTM planning
  7. 07reflection and improvement
session · founder-demo/03 customer-discovery
$ airic run founder-demo --stage customer-discovery
→ loading pack founder-demo
→ role interviewer · process discovery-loop
document
discovery/2026-w26-interviews.md
→ drafted 8 questions from precedents/jtbd-framing.md
→ awaiting review …
approved · committing patch
// 05 — reflection

Self-improvement you can inspect.

Airic does not silently update hidden memory. After work is done, Airic reflects on the process — and proposes changes as Markdown diffs you review and approve.

  • ?What went wrong?
  • ?Which context caused the mistake?
  • ?Was the process unclear?
  • ?Was the document type missing a requirement?
  • ?Should this human judgment become a precedent?
  • ?Should the pack be improved?
Airic learns by editing methodology documents —
with human approval.
patch · precedents/jtbd-framing.md
        @@ -12,6 +12,11 @@
   When framing a customer problem,
   always restate it as a Job-To-Be-Done
   before generating solution candidates.

+ ## Precedent (added 2026-06-27)+ If the user describes a workflow,+ extract the trigger, the desired+ outcome, and the current workaround+ before proposing any solution.- // TODO: revisit framing rules
      
+5 −1 · 1 file
// 06 — comparison

Not another black-box agent.

Traditional agents
Airic
Hidden prompts
Editable Markdown behavior
One-off chats
Durable executable documents
Skills as isolated procedures
Packs as full methodology systems
Silent memory
Human-approved spec patches
Runtime supervision
Reflection-driven improvement
Tool-first automation
Method-first collaboration
Vendor black box
Open source, workspace-local
// 07 — who it's for

Built for hackers, researchers, and method builders.

Airic is open-source and workspace-native. It works with ordinary files, Markdown specs, and git-friendly workflows — designed for people who want to understand, modify, and extend how their agents behave.

AI-native buildersopen-source hackerssoftware architectsstartup foundersconsultantseducatorsmethodology authorsteams turning SOPs into agent workflows
// 08 — get started

Build agents from methods, not prompts.