Tooling Floor

Claude Code skill that surfaces tech-stack tooling gaps before they become 38-commit cleanup chains.

Why this exists

Every project pays for its tooling floor; once at inception (hours) or later in a retroactive cleanup(days to weeks) chain plus the latent bugs that ship in the gap. Skipping the survey doesn't save time; it converts predictable upfront effort into unpredictable later effort. tooling-floor formalizes that asymmetry into a procedure: two entry-paths, a shared toolbelt matrix, and a 4-tier prioritization framework that decides what gets wired before the first business-logic commit and what gets sequenced as multi-version chains afterward.

What it does

• Inventories every language, runtime, framework, and build system in the project
• Maps each tech to the canonical toolbelt matrix (lint / format / test / security / other)
• Classifies every MISSING tool into CRITICAL / HIGH / MEDIUM / LOWER tiers
• Sequences adoption by risk × retro-fit cost (retroactive) or CRITICAL-day-1 (inception)
• Hands off each adopted tool as its own multi-version drain chain, no half-wired tooling

Quick install

npm (recommended):

npx tooling-floor install
# → copies the skill into ~/.claude/skills/tooling-floor/

Manual:

git clone https://github.com/robigewii/tooling-floor ~/.claude/skills/tooling-floor

Usage

Invoke from any Claude Code session:

/tooling-floor

Or via the Skill tool. Example interaction:

> /tooling-floor
Skill: I'll survey this project's tooling floor. First: is this an
inception (fresh project / new tech being added) or retroactive
(existing project, never surveyed) run?
> retroactive
Skill: Inventorying tech stack from package.json + Cargo.toml + ...

Two entry-paths

Inception - fresh projects or new tech being added to an existing stack. Wire CRITICAL gates BEFORE any business logic exists; plan HIGH as v0.x phases; backlog MEDIUM + LOWER with explicit triggers. See SKILL.md § Inception path.

Retroactive - existing projects that were never surveyed, or stacks that drifted past their last survey. Sequence by risk × retro-fit cost; CRITICAL × LOW first (the markdownlint + actionlint + typos + gitleaks quadrant); HIGH × HIGH last. See SKILL.md § Retroactive path.

Supported tech stacks (v1.0)

The toolbelt matrix ships with these stacks:

• TypeScript / JavaScript / Node
• Rust
• C# / .NET
• Python
• Java / Kotlin
• Go
• SQL - T-SQL specific
• SQL - generic / PostgreSQL
• CSS / SCSS
• Docker / containers
• Shell
• Tauri / Electron / native-bridge frameworks
• Cross-cutting (any project)

The matrix grows via community PR -- see CONTRIBUTING.md.

Companion docs

| Doc | Purpose | |---|---| | docs/linter-output-triage.md | Calibration playbook for any new linter - autofix-discipline, per-site disable syntax, rationale comments | | docs/project-wisdom-template.md | Template for capturing project-specific tooling wisdom as it accrues | | docs/case-studies/2026-04-30-eslint-marathon.md | The real 38-commit retroactive ESLint adoption marathon that motivated this skill |

Contributing

Most contributions are stack-row PRs: adding a new language/tech row to the toolbelt matrix. See CONTRIBUTING.md for the procedure (open an issue first, then PR).

Methodology refinements are rarer and require an issue + discussion before PR.

License

MIT see LICENSE.

Acknowledgments

Born from a real 38-commit retroactive ESLint adoption marathon. The asymmetry insight is field-proven, not theoretical.