Devin already reads Agent Skills. Give it one that stops overpaying.

How this maps onto Devin

Agent Skills for the policy, AGENTS.md for what's already always-on.

Devin CLI and Desktop support Agent Skills at .devin/skills/<name>/SKILL.md (project-scoped) or a user-level directory — ~/.config/devin/skills/ on macOS/Linux, %APPDATA%\devin\skills\ on Windows — per Devin's own skills docs. Devin also reads a root AGENTS.md automatically and treats it as always-on; Devin's docs specifically recommend Skills over piling everything into Rules for anything task-conditional, which a routing policy is. That makes Agent Skills the closer fit for Undercut — see Install below for the exact path.

Per Devin's own rules docs, Devin also auto-reads CLAUDE.md, the legacy .windsurfrules file, and — for cross-tool compatibility — .cursor/rules/*.md, .windsurf/rules/*.md, and .claude/ directory files if present in the repo. So if you've already put the policy in front of Claude Code, Cursor, or Windsurf in the same repo, Devin may pick it up with nothing extra installed. Worth checking before you add a second copy.

Undercut's tier names — cheap, standard, frontier, apex — stay generic on purpose. Devin runs its own orchestration across a set of underlying models rather than exposing one vendor's fixed lineup the way a single-model host does, so models.md gives a rule of thumb here rather than a fixed mapping: cheap resolves to the fastest/lightest path Devin has available, standard to its default working behavior, frontier to the highest-reasoning path, and apex to a single batched tie-break call. Nothing here sits between Devin and its model backend — it's a skill Devin reads at dispatch time, no proxy, no network hop.

Install for Devin

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

Option 1 — copy the skill into Devin's skills directory (recommended)

Clone or download undercutsh/firstpass, then copy skills/firstpass/ into .devin/skills/firstpass/ at the project root — the folder Devin's own docs list for project-scoped Agent Skills.

Option 2 — install once, for every repo

Devin also reads a user-level skills directory, so it doesn't need to live in each project: ~/.config/devin/skills/ on macOS/Linux, %APPDATA%\devin\skills\ on Windows.

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. 3Kick off a small Devin session against a repo with the skill installed and check whether it references the routing policy before starting mechanical work

Zero lock-in. It's a policy Devin follows — remove the skill 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 — and Devin's own reliability at scoring the flags hasn't been separately measured the way Claude Code's has (see above); treat that as unverified until tested. Doesn't promise a dollar figure for your workload, and the AGENTS.md/CLAUDE.md auto-pickup described above is documented behavior we haven't independently confirmed end-to-end. Full list of general limits, with the finance/procurement note and the paid-tier disclosure, is on the main page.