Continue already reads a rules folder. Give it one that stops overpaying.

How this maps onto Continue

.continue/rules/ is the confirmed, auto-discovered path.

Continue's documented mechanism is a project-level .continue/rules/ folder at the workspace root: any Markdown file dropped in there is auto-discovered when Continue starts and concatenated into the system message for Agent, Chat, and Edit requests alike — no config-file edit required (per Continue's own rules docs). Each rule file carries YAML frontmatter — name (required), plus optional description, globs, regex, and alwaysApply — and files load in lexicographical order, so a numeric prefix (01-…, 02-…) controls sequencing when you're running more than one. Undercut's SKILL.md already ships a name: and description: frontmatter block for other clients, which happens to satisfy Continue's required fields as-is — no rewrite needed, just a copy.

Continue's config.yaml also supports referencing a rule with a rules: block (uses: author/rule-name for a published Continue Hub block, or uses: file://path/to/rule.md for a local file) — a second, more explicit path if you'd rather wire it through config than rely on folder auto-discovery. Both read the same Markdown file; Install below covers the folder path, since it needs no config edit at all.

Undercut's tier names — cheap, standard, frontier, apex — are generic on purpose. Continue is explicitly bring-your-own-model — you wire up any provider (Anthropic, OpenAI, local Ollama models, and more) in config.yaml yourself, so there's no single fixed lineup to map onto. models.md gives the rule of thumb instead of a fixed slug: cheap resolves to the smallest/fastest model in whatever provider lineup you've configured, standard to your default work model, frontier to the highest-reasoning model you have access to, and apex is reserved for the single batched tie-break call when the ladder is exhausted.

Nothing about this sits between Continue and whichever model backend you've configured. It's plain markdown text Continue reads into its system message, same as any other rule — no proxy, no network hop, nothing to route around if you delete the file.

Install for Continue

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

Option 1 — drop it in .continue/rules/ (recommended, confirmed path)

Run this from the repo root. Continue auto-discovers every Markdown file under .continue/rules/ and loads it at session start — no config.yaml edit needed.

Option 2 — reference it from config.yaml

If you'd rather wire it through config than rely on folder auto-discovery, add a rules: entry pointing at a local file — per Continue's own config reference, this is the same uses: file://… syntax used to pull a published Continue Hub rule block.

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. 3Open Continue's rules panel in the IDE and confirm firstpass shows up as an active rule before starting mechanical work

Zero lock-in. It's markdown text Continue reads — delete the file 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 Continue'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. It also doesn't pick a model for you: Continue is bring-your-own-model, so resolving each tier to a concrete provider and model is still a choice you make in Continue's own config.yaml. And it doesn't promise a dollar figure for your workload. Full list of general limits, with the finance/procurement note and the paid-tier disclosure, is on the main page.