SLM CLI Agent
Lightweight shell helper that translates natural language commands to safe, platform-specific shell scripts and executes them locally on CPU with built-in command security constraints.
02 / Standalone OS Harness Installers, CLI & Python SDK
In SLMAgents 2.0, SLM CLI Agent executes inside the 10-Plane Pure-CPU Agent Harness (slm_harness). Download a self-extracting Individual Agent + Harness OS Installer for macOS, Windows, or Linux (bundling slm_cli_agent + slm_harness + the ./slm CLI), install the full 26-agent desktop suite, or install via Python pip:
1. Individual SLM CLI Agent + 10-Plane Harness Installers (All OSes)
slm_cli_agent + slm_harness + ./slm CLIslm_cli_agent + slm_harnessslm_cli_agent + slm_harness + ./slm CLI2. Full 26-Agent Studio & Desktop OS Installers
3. 10-Plane Harness CLI (./slm) & Python pip Installation
# Option A: Install via PyPI or editable source with the 10-Plane Harness $ pip install slm-cli-agent $ pip install -e ./slm_harness -e ./slm_cli_agent # Option B: Run directly through the Unified 10-Plane Pure-CPU Harness CLI (./slm) $ ./slm ask "Find all .log files modified in the last 7 days and count ERROR lines" $ ./slm chat # Interactive Rich TUI with 18 slash commands (/trace, /models, /arena, /undo) $ ./slm serve # 100% Offline Web AI Studio + OpenAI /v1 API at http://127.0.0.1:7860
03 / 10-Plane Pure-CPU Harness Integration (slm_harness)
Rather than executing as an isolated prompt wrapper, SLM CLI Agent is natively orchestrated by the SLMAgents 2.0 10-Plane Pure-CPU Harness (via slm_harness.runtime.legacy_bridge and CascadeRouter). Every call benefits from hybrid llama.cpp GGUF (Q4_K_M) + Quantized INT4/INT8 CPU execution, watermarked LRU memory governance, grammar-constrained decoding, and deterministic verification at $0 API cost:
| Harness Plane | Subsystem | How It Powers SLM CLI Agent on Pure CPU |
|---|---|---|
| Plane 6 | CascadeRouter (<35ms) |
Stage 2 (POSIX / Shell Intent Signature) & Stage 3 (1024-d Embedding Margin) |
| Plane 2 & 5 | MemoryArena & Hybrid CPU Runtime |
Specialist Role: coding (Qwen2.5-Coder-1.5B INT4 / Qwen3-4B GGUF Q4_K_M). Managed inside Watermark LRU MemoryArena (lite 2.6 GB / balanced 4.7 GB / power 8.2 GB) with zero-copy session sharing and automatic OOM tier step-down (8B -> 4B -> 3B -> 0.8B). |
| Plane 7 & 8 | Deterministic Verifier & Governance | Plane 8 Governance Risk Lattice (L0_PURE to L4_SYSTEM_OR_NET), Destructive Pattern Blocker, POSIX rlimit Sandbox & ApprovalBroker |
| Plane 1, 3 & 4 | Kernel, Artifact Scratchpad & 4-Tier Memory | SQLite WAL CheckpointStore (/fork, /rewind, /undo), content-addressed artifact://<sha12> handles to prevent prompt bloat, and hybrid SQLite FTS5 BM25 + HNSW vector memory. |
04 / Source Checkout (Git Sparse-Checkout)
Clone this agent alongside the 10-Plane Pure-CPU Harness (slm_harness) from the monorepo using Git sparse-checkout — no need to download all 26 agents:
Option 1 — Sparse Checkout with 10-Plane Harness (Recommended)
# 1. Create and enter a new directory $ mkdir slm_cli_agent_workspace && cd slm_cli_agent_workspace # 2. Initialise empty git repo and add remote $ git init $ git remote add origin https://github.com/t00114218-stack/SLMAgents.git # 3. Enable sparse-checkout for both slm_harness and slm_cli_agent $ git sparse-checkout init --cone $ git sparse-checkout set slm_harness slm_cli_agent # 4. Pull the 10-Plane Harness and agent source $ git pull origin main
Option 2 — Full Repository Clone
$ git clone https://github.com/t00114218-stack/SLMAgents.git $ cd SLMAgents $ ./slm doctor
Note: After checkout, install the harness and agent in editable mode with pip install -e ./slm_harness -e ./slm_cli_agent or download the pre-packaged OS Harness Installer above.
05 / Configuration & API Reference
Constructor Parameters
| Parameter | Type / Default | Description |
|---|---|---|
| model_path | str | None | Explicit path to GGUF (Q4_K_M) or quantized ONNX model directory. Defaults to zero-copy sharing via the 10-Plane Harness MemoryArena. |
| cache_dir | str | None | HF model directory path. Also settable via SLM_CLI_AGENT_CACHE_DIR. |
| n_ctx | int | 2048 | Context window size in tokens. Also settable via SLM_CLI_AGENT_N_CTX. |
| n_threads | int | 4 | CPU threads for the 10-Plane Pure-CPU Harness (auto-tuned to physical P-cores). Also settable via SLM_CLI_AGENT_N_THREADS. |
| system_prompt | str | None | Optional custom system prompt instructions overriding the default template. |
| user_input | str | None | Optional additional user-supplied target parameters or variables. |
generate_command() Parameters
| Parameter | Type / Default | Description |
|---|---|---|
| query | str | Required. Natural language description of what you want to achieve. |
| stream | bool | False | If True, enables token streaming of the model's explanation and thinking process. Returns a Python generator. |
| system_prompt | str | None | Optional custom system prompt instructions overriding the default template. |
| user_input | str | None | Optional additional user-supplied target parameters or variables. |
from slm_cli_agent.cli_agent import SLMCLIAgent
agent = SLMCLIAgent()
# Generate command from query
cmd, explanation = agent.generate_command("find all files ending with .py in current folder")
print(cmd)
# Output: find . -name "*.py"
Unified Command Execution (run())
For convenience, you can translate and execute shell instructions in a single method call:
result = agent.run("list directory contents in clean format")
print(f"Executed Command: {result['command']}")
print(f"Success: {result['success']}")
print(f"Output:\n{result['stdout']}")
Safe Command Execution
The CLI Agent provides a safe command execution API that screens script blocks for destructive patterns prior to invoking them in the local environment.
code, stdout, stderr = agent.execute_command("echo 'Hello CLI Agent'")
print(f"Return code: {code}")
print(f"Stdout: {stdout}")
rm -rf /, mkfs, dd if=, shutdown) are caught by the protection engine, resulting in a return code of -1 and an error returned to prevent command-line accidents.
Complex Command Pipelines
The CLI Agent handles complex natural language requirements involving multiple nested operations, pipes, file redirections, and limits:
query = (
"Find all files in /var/log ending with .log modified in the last 7 days, "
"search them for the term 'ERROR' case-insensitively, count the matching "
"lines for each file, sort the count descending, and output the top 5 records "
"while ignoring permission denied errors."
)
result = agent.run(query)
print(result["command"])
find /var/log -type f -name "*.log" -mtime -7 2>/dev/null | xargs grep -io "ERROR" 2>/dev/null | cut -d: -f1 | sort | uniq -c | sort -rn | head -n 5