https://github.com/ckpxgfnksd-max/prompt-compressor-openclaw
PassAudited by VirusTotal on May 11, 2026.
Overview
Type: OpenClaw Skill Name: prompt-token-saver Version: 0.2.0 The 'prompt-token-saver' skill is a utility designed to reduce LLM token consumption by enforcing concise communication and data logging. It achieves this through a JavaScript hook (handler.js) that uses regex to strip filler words from user input and by providing instructions (soul-snippet.md) to be appended to the agent's core SOUL.md file. The code and instructions are transparent, align with the stated purpose of token efficiency, and show no signs of malicious intent, data exfiltration, or harmful prompt injection.
Findings (0)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Future memories and logs may be shorter and less nuanced, which can save tokens but may lose detail.
The skill intentionally changes how persistent memory and log files are written, which can affect future context and summaries.
You save tokens on every response and in all persistent files... When writing to MEMORY.md or memory/YYYY-MM-DD.md
Install only if you want persistent concise-memory behavior; keep a backup of SOUL.md and remove the snippet if it interferes with accuracy or recall.
If enabled, the hook may alter the wording of user prompts before the model sees them, which could affect intent in cases where exact phrasing matters.
The included handler is disclosed as a hook that can run automatically on incoming messages if the user manually installs and enables it.
Runs as a message:received hook — compresses user messages BEFORE the LLM sees them.
Use the hook only when you are comfortable with automatic prompt rewriting; avoid it for legal, safety-critical, or exact-wording tasks.