A reliable file-writing workflow skill for AI agents designed to eliminate JSON truncation and encoding errors when handling large files or CJK characters.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install llm-safe-write
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install llm-safe-write using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is llm-safe-write?
The llm-safe-write skill provides a robust, incremental strategy for writing large files or content containing CJK (Chinese, Japanese, Korean) and special characters. Standard AI file writing tools often fail due to transport-layer JSON truncation when multi-byte characters cross internal buffer boundaries, leading to compilation failures or incomplete code blocks. This skill mitigates that vulnerability entirely.
By leveraging Openclaw Skills, AI agents adopt a proactive skeleton and target-edit approach rather than executing direct, high-risk writes. It guarantees technical accuracy and reliability across long source files, complex configuration schemas, and multi-lingual data payloads without losing critical text or breaking execution environments.
llm-safe-write Use Cases
- Writing source code or technical documentation files that exceed 50 lines.
- Injecting Chinese, Japanese, Korean, or specialized non-ASCII strings directly into application files.
- Recovering immediately after an agent's standard file tool fails with Unterminated string or truncation errors.
- Initializing multi-function scripts in Python, JavaScript, or TypeScript without risk of silent transport corruption.
How llm-safe-write Works
- File Evaluation: The agent assesses the file length and character encoding before writing. If it exceeds 50 lines or contains non-ASCII characters, it routes to this specialized skill workflow.
- Skeleton Construction: The agent creates an ASCII-only file framework under 50 lines, swapping complex content and CJK strings with highly unique, descriptive placeholders.
- Initial Structural Write: The agent writes the lightweight skeleton file securely to the storage environment.
- Incremental Content Injection: The agent invokes targeted, short edit operations changing fewer than 30 lines at a time to replace placeholders with real code or multi-byte text.
- Post-Write Verification: A spot-check reads the boundaries of the file to verify completeness, reporting success concisely without leaking internal tool operations.
llm-safe-write Setup
To deploy this module within your Openclaw Skills runtime environment, verify that your active agent configuration includes the necessary file manipulation permissions for reading, writing, and editing.
No standalone terminal setup is needed, but the target file directories must exist beforehand to prevent filesystem errors:
mkdir -p /path/to/target/directory
llm-safe-write Data Schema & Taxonomy
The skill manages tracking and insertion operations across file formats by standardizing layout boundaries with strict code-comment anchors:
| File Type | Anchor Taxonomy Structure | Live Implementation Example |
|---|---|---|
| Python | # PLACEHOLDER_name |
pass # PLACEHOLDER_process_data |
| JavaScript / TypeScript | // PLACEHOLDER_name |
// PLACEHOLDER_render_view |
| JSON / Data Schemas | Same-type valid literal defaults | "meta_title": "PLACEHOLDER_seo" |
This deliberate schema ensures that files maintain formatting validity and remain parseable throughout the incremental update cycle.
llm-safe-write Advanced Features
- Multi-Format Architectural Parsing: Context-aware workflows tailored specifically for Python classes, JS/TS functions, structured YAML configurations, and Markdown sections.
- Automated Collision Recovery: Implements contextual expanding and precise replacement hooks to automatically bypass error types such as matching ambiguity or stale file views.
- Stable Boundary Anchoring: Utilizes last-line tracking replication algorithms to append code cleanly onto massive legacy files without risking a full-file transport reserialization error.
SKILL.md
Loading
Related Openclaw Skills
Gateway Upgrade Local Fork
An end-to-end upgrade tool for local OpenClaw gateway services that automates backups, systemd user environment merging, and serial GPU index rebuilds.
Huawei Cloud CLI Guidance
A complete, step-by-step developer guide to installing, configuring, and automating operations using the Huawei Cloud CLI (KooCLI) command-line interface.
Xiaohongshu Crawler
An advanced crawling and reporting tool to scrape, filter, and analyze trending Xiaohongshu notes based on keywords and engagement metrics.
Stock Analysis
An advanced market intelligence and portfolio management toolkit for AI agents to analyze stocks, crypto, and market rumors.
BasicOps MCP Operator
An AI agent skill designed to safely execute day-to-day BasicOps tasks, updates, and reviews via a Model Context Protocol (MCP) server.
Imans CLI Integration
An AI-agent skill that connects OpenClaw to the Imans CLI, enabling seamless queries of workspace metadata, catalogs, and sales orders.
Featured*
