Droid already has a native Skills folder. Give it one that stops overpaying.

How this maps onto Factory

.factory/skills/ is Droid's documented, first-class Skills mechanism.

Per Factory's own Skills reference, Droid discovers Skills from a defined precedence of locations — a folder-specific .factory/skills/ nearest the working directory, the project-level .factory/skills/ at the repo root, a personal ~/.factory/skills/, installed plugins, built-ins, and mission-scoped skill folders, with legacy compatibility paths at .agents/skills/ and .agent/skills/. Each skill is a directory containing an entry point that must be named exactly SKILL.md — the same filename and frontmatter shape (name, description, plus optional fields like disable-model-invocation) that Undercut already ships. Droid reads every skill's name and description up front, matches them against what you're asking it to do, and loads the full body only when a skill applies — the same progressive-disclosure model as Claude Code's own Skills.

That's a genuinely good fit: of every client this site covers, Factory is one of the few with a documented, native SKILL.md-shaped discovery path rather than a repurposed rules file. Factory separately documents AGENTS.md discovery — Droid also walks up to the git root checking .factory/, .agents/, and .agent/ directories (plus personal equivalents in your home folder) for an always-on rules file named AGENTS.md, agents.md, Agents.md, or (for compatibility) CLAUDE.md/Claude.md — but Factory's own docs draw the line clearly: AGENTS.md is for "always-on repository instructions, commands, conventions, and safety rules," while a Skill is a "reusable workflow Droid can choose or the user can invoke." Undercut's routing rubric is a reusable workflow, not a standing house rule, so the Skills path is the better-fitting home for it — with an AGENTS.md append as a belt-and-suspenders option below.

Undercut's tier names — cheap, standard, frontier, apex — are generic on purpose. Factory is model-agnostic, letting you pick from several provider models per session, so models.md gives a rule of thumb rather than a fixed slug: cheap resolves to the smallest/fastest model your Factory workspace has enabled, standard to the default work model, frontier to the highest-reasoning production model, and apex to a single batched tie-break call. Nothing here sits between Droid and its model backend — it's plain markdown text Droid reads into context, no proxy, no network hop, nothing to route around if you remove the folder.

Install for Factory

Free. MIT. Nothing installs, nothing proxies your traffic.

Option 1 — the skills CLI picker (fastest)

skills.sh lists Droid as a supported install target. Run this from your project root and pick Droid when the picker asks which agent to install for:

Option 2 — copy it by hand (confirmed path)

Directly confirmed against Factory's Skills docs: the folder name must match the name field in SKILL.md (firstpass), and the entry point must be named exactly SKILL.md.

Prefer it available in every repo instead of just this one? Copy the same folder to ~/.factory/skills/firstpass (personal scope) rather than the project's .factory/skills/.

Option 3 — also append to AGENTS.md

Belt-and-suspenders: if you'd rather the rubric read as a standing rule than a workflow Droid has to choose to invoke, append the same text to the project's AGENTS.md, one of the files Droid's documented discovery walk reads automatically from the repo root.

Validate it yourself

1. 1Read the published results — testing/README.md and the raw JSON in testing/results/
2. 2Reproduce for free, no API key, no spend: node src/main.js --mock
3. 3Ask Droid to run /firstpass (or describe a batch of mechanical work) on one small task and check which tier it names — a single smoke test is pennies

Zero lock-in. It's a folder Droid reads — delete it and you're back to status quo.

Six flags assign a tier. Three triggers are the only way up.

This is a named, versioned spec — not internal logic we could quietly change. When the rubric changes (a new flag, a different cap), the version number moves and the changelog says why.

Every unit of work gets scored against six flags, then assigned the cheapest tier that can plausibly handle it. It only moves up a tier when something objective proves it needs to — a failed check, a measured disagreement, an explicit uncertainty flag. It never moves up on a guess, and it never moves back down mid-task.

 unit of work
      |
  6-flag rubric
      |
  +---+----+
  | 0 flags|--> cheap
  | 1-2    |--> standard
  | 3+/own |--> frontier
  +---+----+
      | fail x2 / disagree / uncertain
      v
  escalate +1 tier
  (residue only)
      |
  still unresolved?
      v
  single batched
  apex tie-break

Why it holds up

Flags steer. Verification and escalation decide.

Even when the rubric flags are scored imperfectly, the escalator still routes work to the right tier almost every time — because a wrong flag only ever costs one extra cheap attempt, never a wrong answer or a big bill.

Honest limits

What this doesn't do.

Not a proxy, not a gateway, doesn't auto-flag in production (Droid scores the flags itself — measured robust to getting them wrong, see above), and doesn't promise a dollar figure for your workload. It also doesn't prove your number until you run it — every figure here is an observation from our tasks and our graders, not a guarantee about yours. We've confirmed .factory/skills/ and its SKILL.md shape against Factory's own docs, but we haven't run the npx skills add picker or a live Droid session ourselves end-to-end, and we haven't verified how reliably Droid self-invokes this skill versus waiting to be asked — see the install callout above for the specifics. Full list, with the finance/procurement note and the paid-tier disclosure, is on the main page.