version: "1.0.0"

Context Capsule

⚠️ MOVED: canonical home is

openclaw-skills/skills/context-capsule.

This copy is frozen.

Compress agent session history ~99% before it hits the LLM, so you stop paying to re-send the whole transcript on every call. Works with any model — Claude, GPT, Ollama, Mistral, LM Studio.

Self-contained (v1.4.0): The compression core is bundled directly in this

skill (src/compression.ts). There is no external runtime dependency, and

the plugin makes no network, file-system, or on-chain calls. Everything

runs locally using only Node's built-in zlib and crypto.

Data handling: Every message is scanned by an inline vault gate before it

reaches the compression core or the model — on every code path, including short

sessions, the verbatim tail, and the compression-failure fallback. API keys,

tokens, credentials, PII, and card numbers are replaced with typed

placeholders. No matched values are logged, only category counts. Redaction is

best-effort pattern matching, not a guarantee — do not rely on it as the

sole protection for highly sensitive chats.

When to use

• Long-running agent sessions (default: more than 20 messages) where the

transcript is large and you want to cut per-call token cost.

• Any model/provider — local or hosted.

When NOT to use

• Sessions that require exact, verbatim transcript fidelity. Older history is

summarized into a compact capsule; detail and nuance can be lost. Only the most recent 10 messages are kept verbatim.

• As your only safeguard for secrets/PII. The vault scan is best-effort, and the

compressed history is injected into the system context position.

How it works

Keeps the last 10 messages verbatim. Replaces everything older with a short pointer string (zlib ratio, topic words, Merkle prefix). The pointer carries no facts from the older turns and this copy has no retrieval, so older details are not available to the model. No model-answer accuracy has been measured for it.

Savings

On the 109-message fixture in packages/context-capsule/bench/fixtures, the same pointer format built by the standalone library over the whole session is 53 estimated tokens (chars/4) where the full JSONL history is 7,919. That is the initial prompt size only; it says nothing about whether the model can still answer questions about the older turns.

Install

This skill is self-contained — no extra packages to install. Register it as your context engine in openclaw.json:

{
  "plugins": {
    "slots": { "contextEngine": "context-capsule" }
  }
}

Optional config (defaults shown):

{
  "plugins": {
    "entries": {
      "context-capsule": { "minMessages": 20, "keepRecentMessages": 10 }
    }
  }
}

Source

github.com/Parad0x-Labs/dna-x402/tree/main/packages/openclaw-context-capsule

The standalone library (@parad0x_labs/context-capsule) is published separately on npm for non-OpenClaw use; this skill vendors only the two pure functions it needs and does not depend on it at runtime.

💜 If Context Capsule is quietly shrinking your token bill, a ⭐ on ClawHub helps other agent builders find it.