本文へ移動
cccskills
無料GitHub で公開

rtk-exec

Execute CLI commands through RTK proxy for token-optimized output

インストール方法を見る

含まれるファイル(2)

  • SKILL.md5.8 KB
  • scripts/rtk-wrapper.cjs9.7 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

RTK Exec Skill

Execute CLI commands through the RTK (Rust Token Killer) proxy to reduce LLM token consumption by 60-90%. RTK is a CLI command proxy — it wraps existing shell commands and compresses their output using smart filtering, grouping, truncation, and deduplication.

Important: RTK is NOT an AI prompt tool. It is a CLI output compressor. You pass it a regular shell command and it returns compressed output.

Options

<command> [args...]   Required. CLI command to proxy through RTK (e.g., "ls .", "git status", "cargo test")
--gain               Show token savings statistics (rtk gain)
--version            Show RTK version (rtk --version)
--init               Initialize RTK for current project (rtk init -g)
--timeout <ms>       Execution timeout (default: 120000, max: 600000)
--working-dir <dir>  Working directory for command execution

Workflow

1. Pre-checks
   - Verify `rtk` binary is installed (which rtk)
   - No authentication required
2. Build command
   - Special: rtk gain | rtk --version | rtk init -g
   - Proxy: rtk <command> [args...]
3. Execute
   - Run via Bash tool with timeout (default 2min, max 10min)
   - Or use helper script: node .claude/skills/rtk-exec/scripts/rtk-wrapper.cjs
4. Parse output
   - RTK always returns plain text compressed output (no JSON modes)
   - Capture stdout and stderr
5. Report results
   - Format output with execution metadata and token savings info

Supported Commands

RTK supports 100+ CLI commands across categories:

CategoryCommands
File opsls, find, grep, cat, tree, du, diff
Gitgit status, git log, git diff, git push, git pull, git branch
Test runnerscargo test, pytest, vitest, jest, go test, bun test
Linterseslint, tsc, ruff, clippy, golangci-lint, rubocop
Package managerspnpm install, pip install, cargo build, npm install
Containersdocker build, docker ps, kubectl get, kubectl logs
Build toolsmake, cmake, gradle, mvn
Systemps, top, df, netstat, env

Output Format

Success

[RTK Exec] Completed

Command: rtk git status
Duration: 0.3s
Working Dir: /path/to/project

--- Output ---
{rtk compressed output}

Success (--gain)

[RTK Exec] Token Savings

--- Gain Report ---
{rtk gain statistics showing tokens saved per command}

Failure

[RTK Exec] Failed

Command: rtk cargo test
Error: {error_message}
Exit Code: {code}
Stderr: {stderr}
Suggested Fix: {suggestion}

Helper Script

For complex executions or programmatic use:

node .claude/skills/rtk-exec/scripts/rtk-wrapper.cjs --command "cargo test" [options]

The wrapper provides:

  • Binary availability check
  • Safe command construction
  • Timeout handling with graceful SIGTERM → SIGKILL escalation
  • Structured JSON output for programmatic consumption

Examples

# List files with compression
/rtk-exec "ls -la src/"

# Git status (often 70%+ token reduction)
/rtk-exec "git status"

# Run Rust tests
/rtk-exec "cargo test"

# Run Python tests
/rtk-exec "pytest tests/ -v"

# TypeScript type check
/rtk-exec "tsc --noEmit"

# Show token savings stats
/rtk-exec --gain

# Initialize RTK for project
/rtk-exec --init

# With timeout override
/rtk-exec "cargo build --release" --timeout 300000

# Specify working directory
/rtk-exec "git log --oneline -20" --working-dir /path/to/repo

Integration

Works with the orchestrator pattern:

  • Main conversation delegates CLI execution via this skill
  • Results are returned to the main conversation for further processing
  • Particularly effective for test runs, lint checks, and git operations where output is verbose
  • Can be chained: rtk-exec for output collection → Claude expert for analysis

Availability Check

rtk-exec requires the RTK binary to be installed. The skill is only usable when:

  1. rtk binary is found in PATH (which rtk succeeds)
  2. No authentication or API keys required

If the binary check fails, this skill cannot be used. Fall back to direct Bash tool execution.

Note: This skill is invoked via /rtk-exec command or delegated by the orchestrator. It is most useful for any task that produces verbose CLI output that would otherwise consume large amounts of context tokens.

Agent Teams Integration

When used within Agent Teams:

  1. As delegated task: orchestrator explicitly delegates CLI execution for token-efficient output
  2. Hybrid workflow: Claude team member plans → rtk-exec runs commands → Claude analyzes compressed output
  3. Batch execution: Multiple rtk-exec invocations in parallel for different commands
Orchestrator delegates CLI task
  → /rtk-exec invoked with command
  → Compressed output returned to orchestrator
  → Analyst processes compressed result
  → Iterate if needed

Token Savings

RTK uses four compression strategies:

StrategyDescriptionTypical Savings
Smart FilteringRemoves redundant/noise lines based on command type30-50%
GroupingCollapses repeated patterns into summary counts20-40%
TruncationClips excessively long lines with ellipsis10-20%
DeduplicationRemoves identical repeated output blocks15-30%

Combined effect: 60-90% token reduction on typical CLI output (git log, test results, lint reports).

Installation

# macOS via Homebrew
brew install rtk

# Universal install script
curl -fsSL https://raw.githubusercontent.com/rtk-ai/rtk/refs/heads/master/install.sh | sh

# From source via Cargo
cargo install --git https://github.com/rtk-ai/rtk

# Verify installation
rtk --version

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

Pre-action boundary checking — validates agent tool calls against declared capabilities and task contracts

日本語の概要は準備中です。原文の説明を表示しています。

baekenough/second-brain152026年10月8日 更新

Auto-detect project context and optimize harness — deactivate unused agents/skills, suggest missing experts, generate project profile

日本語の概要は準備中です。原文の説明を表示しています。

baekenough/second-brain152026年10月8日 更新

Adversarial code review using attacker mindset — trust boundary, attack surface, business logic, and defense evaluation

日本語の概要は準備中です。原文の説明を表示しています。

baekenough/second-brain152026年10月8日 更新

Apache Airflow best practices for DAG authoring, testing, and production deployment

日本語の概要は準備中です。原文の説明を表示しています。

baekenough/second-brain152026年10月8日 更新

Alembic migration patterns for naming conventions, safety checks, expand-contract, env.py configuration, and CI integration

日本語の概要は準備中です。原文の説明を表示しています。

baekenough/second-brain152026年10月8日 更新

Pre-routing ambiguity analysis — scores request clarity and asks clarifying questions when needed (inspired by ouroboros)

日本語の概要は準備中です。原文の説明を表示しています。

baekenough/second-brain152026年10月8日 更新

baekenough のスキルをすべて見る

このスキルの問題を報告する