framework-field-guide

Framework Field Guide

A grounded, agent-callable catalog of ~43 corporate operating frameworks — RACI, OKRs, Theory of Constraints, DORA, Wardley Mapping, JTBD, Kanban, and the rest — plus a Claude Code skill that routes a decision to the right framework and warns you when two of them conflict.

It is a router, not a lecture. The catalog holds the knowledge (origin, structure, SWOT, a real company case, and how each framework combines or conflicts with the others). The frameworks-thinking skill decides which framework fits the decision in front of you, names the layer it lives in, and pulls just that framework's file on demand.

The differentiator is grounding honesty: every factual claim is labeled with where the evidence actually comes from. Owner-reported numbers say so. Claims that couldn't be grounded independently were cut or flagged — not dressed up.

What's inside

skills/frameworks-thinking/
├── SKILL.md                    # the router (progressive-disclosure logic)
└── references/
    ├── catalog-pointer.md      # the catalog load table + ID convention
    └── catalog/                # the grounded catalog (the knowledge)
        ├── OVERVIEW.md         # the reasoning layer — 4-layer model, tensions, combine/conflict maps, fad graveyard
        ├── catalog.json        # machine index: framework ID → path + facets
        ├── bibliography.md     # 57 resolvable citation URLs, tiered by authority
        ├── gap_audit.md        # what's deliberately NOT in v0.1 yet (route here for "is there a framework for X?")
        ├── operating/          # "how do we RUN the org" — strategy / alignment / execution / measurement (42 files)
        ├── thinking/           # "how do we REASON" — mental models (Cynefin)
        └── domains/            # "how do we decide THIS class" — pricing, negotiation, M&A (reserved taxon)

The catalog covers the operating layer in depth (42 frameworks across strategy, alignment, execution, and measurement). The thinking/ and domains/ taxa are part of the taxonomy and carry what's grounded today; gap_audit.md is the honest map of which clusters are thin, so you always know what the catalog does and doesn't cover.

Install (as a Claude Code skill)

The catalog is bundled inside the skill folder, so installing is one copy:

1. Copy the skills/frameworks-thinking/ directory into your project's .claude/skills/ directory (so you end up with .claude/skills/frameworks-thinking/SKILL.md).
2. That's it. The skill's references/ — including the whole catalog/ — travels with it, and every path inside the skill is relative to the skill root, so nothing points outside the folder.

Claude Code discovers skills in .claude/skills/, so the agent can now invoke frameworks-thinking for decisions like "which framework should I use to decide what to build next quarter?" or "two teams keep stepping on each other — how do we fix ownership?" (Depending on your Claude Code version, it's available as the /frameworks-thinking skill and/or auto-selected by the agent when a routing question comes up.)

You can also just read the catalog directly — skills/frameworks-thinking/references/catalog/OVERVIEW.md is a standalone reference guide with no tooling required.

How to read the evidence

There are two grading systems, and they mean different things.

1. Inline evidence markers — attached to individual claims inside each framework file:

Re-grounding outcome: of the 83 publish-critical claims audited, 33 grounded independently, 32 were labeled self-reported, and 18 were cut, corrected, or downgraded. (That's the audit tally. Separately, the shipped files carry the surviving inline markers: [source:] and [self-reported:] on individual lines, plus [case cut] / [corrected] / [unverified] where re-grounding changed a claim — a different count, because one audited claim can map to more or fewer inline marks.)

2. catalog.json evidence facets — structured metadata on each framework as a whole (not the same as the inline markers):

License

• Code (the skill and any scripts) — MIT.
• Content (the framework catalog) — Creative Commons Attribution 4.0 (CC BY 4.0).

Third-party framework names and the ideas of their originators remain the property of their respective owners; this catalog describes and cites them and reproduces no copyrighted work in full. See LICENSE (code), LICENSE-CONTENT (catalog), and LICENSE_LEDGER.md.

Built by Victor Sowers (STEEPWORKS). Catalog and skill released under the licenses in LICENSE and LICENSE-CONTENT.