cost-booster-route — Claude Code skill: task routing at $0
Sorts a batch of tasks across the router's three tiers and shows how many landed in Agent Booster with no LLM call.
npx skills add ruvnet/ruflo --skill cost-booster-route -a claude-codeWhat cost-booster-route does
The skill wraps mcp__plugin_ruflo-core_ruflo__hooks_route and counts the tasks the router marked as Agent Booster, Tier 1. Those run as WASM transforms, so no model is invoked and the cost is structurally $0. Everything else falls into Tier 2 or 3 and gets an upper-bound estimate.
Reach for it before a batch of lookalike edits, or when cost-report shows Sonnet or Opus spend on work that reads like a plain transform: var-to-const, add-types, add-error-handling, async-await, add-logging, remove-console. Tasks come from the argument or from recent cost-tracking entries pulled with memory_search. Each one goes through the router, and the responses split on whether they carry the literal [AGENT_BOOSTER_AVAILABLE]. A batch is capped at 50.
The report lists task counts per tier, the bypass share and the booster intents it saw, then writes the result to memory_store under the cost-tracking namespace so cost-report can read the tier signal. One caveat matters: the marker only appears when the upstream router fills routeResult.agentBoosterIntent.type, and the semantic VectorDb path in the published CLI does not always fire. Treat the split as a lower bound. The <1ms and $0 figures are structural; the 352× speedup claim is not verified here.
What it helps with
- estimate how many tasks can skip the LLM before a batch run
- spot simple transforms already billed to Sonnet or Opus
- get an upper-bound spend figure for Tier 2 and 3
- feed the tier signal into
cost-reportviamemory_store - break a task list down by booster intent
Example prompts for the agent
- Route the recent cost-tracking tasks through cost-booster-route
- Run cost-booster-route on this: convert var to const across the package
- cost-booster-route --from-recent
How to install cost-booster-route
1Skills CLI (Vercel)
npx skills add ruvnet/ruflo --skill cost-booster-route -a claude-code2Manually: put the files in place
~/.claude/skills/cost-booster-route/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/cost-booster-route/
1Skills CLI (Vercel)
npx skills add ruvnet/ruflo --skill cost-booster-route -a codex2Manually: put the files in place
.agents/skills/cost-booster-route/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add ruvnet/ruflo --skill cost-booster-route -a cursor2Manually: put the files in place
.cursor/skills/cost-booster-route/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install ruvnet/ruflo cost-booster-route2Manually: put the files in place
.github/skills/cost-booster-route/SKILL.md1Skills CLI (Vercel)
npx skills add ruvnet/ruflo --skill cost-booster-route -a gemini-cli2Platform CLI
gemini skills install https://github.com/ruvnet/ruflo --path plugins/ruflo-cost-tracker/skills/cost-booster-routeManually: put the files in place
.devin/skills/cost-booster-route/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add ruvnet/ruflo --skill cost-booster-route -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about cost-booster-route
How do I install cost-booster-route in Claude Code?
Run npx skills add ruvnet/ruflo --skill cost-booster-route -a claude-code. The skill file lands at ~/.claude/skills/cost-booster-route/SKILL.md.
What does cost-booster-route do?
It sends tasks through hooks_route, splits them by the [AGENT_BOOSTER_AVAILABLE] marker, and reports how many bypassed the model at $0 versus how many stayed on an LLM.
Why can the Tier 1 share come out low?
The marker shows up only if the upstream router populates routeResult.agentBoosterIntent.type, and the semantic VectorDb path in the published CLI does not always trigger it. Read the split as a lower bound.
Source text
SKILL.md · MIT
---
name: cost-booster-route
description: Route tasks through hooks_route, partition by Agent Booster availability, and report Tier 1 bypass utilization with $0 cost
argument-hint: "[--from-recent] | <task-description>"
allowed-tools: mcp__plugin_ruflo-core_ruflo__hooks_route mcp__plugin_ruflo-core_ruflo__memory_search mcp__plugin_ruflo-core_ruflo__memory_list Bash
---
# Cost Booster Route
Wraps `mcp__plugin_ruflo-core_ruflo__hooks_route` and reports how many tasks the 3-tier router classified as Agent Booster (Tier 1) eligible. Tier 1 bypasses run as WASM transforms — no LLM call, structurally **$0** cost.
## When to use
Before a batch of similar tasks, or when `cost-report` shows Sonnet/Opus spend on descriptions that look like simple transforms (`var-to-const`, `add-types`, `add-error-handling`, `async-await`, `add-logging`, `remove-console`).
## Steps
1. **Collect tasks** — single arg or recent entries from `cost-tracking` via `memory_search`. Cap batch at 50.
2. **Route each** — call `hooks_route` with the description; capture the full response string.
3. **Partition** — group by whether the response contains the literal `[AGENT_BOOSTER_AVAILABLE]` (Tier 1) vs. not (Tier 2/3). Extract `[TASK_MODEL_RECOMMENDATION] Use model="X"` when present.
4. **Compute spend** — Tier 1 partition: $0. Tier 2/3 partition: per-task upper-bound from REFERENCE.md pricing × recommended model.
5. **Report**
```
=== Booster bypass report ===
Tasks analyzed: 50
Tier 1 (booster): 18 (36%) — $0.00
Tier 2/3 (LLM): 32 (64%) — $X.XX (upper-bound)
Booster intents: var-to-const (8), add-types (5), remove-console (5)
```
6. **Persist** — `memory_store --namespace cost-tracking --key "booster-route-$(date +%Y%m%d-%H%M%S)" --value '{"tier1": N, "tier2_or_3": M, ...}'` so `cost-report` picks up the tier signal.
## Caveats — claimed upstream, not yet verified
- `[AGENT_BOOSTER_AVAILABLE]` fires only when the upstream router populates `routeResult.agentBoosterIntent.type` (`v3/@claude-flow/cli/src/mcp-tools/hooks-tools.ts:1228`). The published CLI's semantic-VectorDb path does not always trigger the classifier — treat the partition as a **lower bound** on Tier 1 eligibility.
- CLAUDE.md root claims `<1ms` latency and `352× faster` than LLM. `<1ms` and `$0` are structural; `352×` is **claimed upstream, not yet verified** here. Report what the router actually returns.
- See `docs/benchmarks/0002-baseline.md` for the full upstream-claims-vs-measured table.
## Cross-references
ADR-0002 Decision #1 · ruflo-intelligence ADR-0001 §"Neutral" (closes the routing-outcomes loop via `cost-optimize` step 8) · CLAUDE.md root §"3-Tier Model Routing (ADR-026)".
© ruvnet · MIT · ruvnet/ruflo
