Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

rlm-controller

bởi skywyze

Web & Frontend Development
866Lượt tải do clawskills.sh công bố
2Sao
0Lượt xem
0 lượt thíchĐăng nhập để thích kỹ năng này
—Chưa có đánh giá
Đánh giá của bạn

Tóm tắt

RLM-style long-context controller that treats inputs as external context, slices/peeks/searches, and spawns.

Mô tả

Kỹ năng này làm gì

A policy-driven scaffold for processing inputs too large for a single context window. Stores input as an external context file, slices or searches it in chunks, and spawns batched subcalls to aggregate structured results. Hard limits on recursion depth, slice count, and subcall count are enforced throughout.

Handles inputs that exceed the model context window by splitting work into bounded subcalls rather than truncating or requiring manual chunking.

Khi nào nên dùng

  • Scanning 50,000-line server logs for error patterns
  • Extracting structured data from large JSONL datasets
  • Analyzing an entire repository across multiple files at once
  • Summarizing dense technical documentation in controlled sections
  • Running keyword searches across multi-file audit trails

Quy trình ví dụ

ĐẦU VÀO

User asks: Find all authentication errors in this 80,000-line application log

AGENT
  1. 1

    Stores the log file as external context via rlm_ctx.py store

  2. 2

    Generates a keyword-based slice plan with rlm_auto.py targeting auth-related terms

  3. 3

    Creates async batches of slices with rlm_async_plan.py

  4. 4

    Spawns subcalls via sessions_spawn to process each batch

  5. 5

    Aggregates subcall results into a structured error report

ĐẦU RA

Structured list of authentication error occurrences with line numbers and surrounding context

Cài đặt

clawhub install skywyze/rlm-controller

Chạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.

Liên kết