LaunchKit · 2026
Back to Skills

chain-of-density

Iteratively densify text summaries using Chain-of-Density technique.

1
1.2k downloads
by @killerapp

Setup & Installation

openclaw skills install @killerapp/chain-of-density

Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:

npx clawhub install chain-of-density

What This Skill Does

Chain-of-Density (CoD) is an iterative summarization method from a 2023 research paper. It compresses text through 5 passes, each identifying missing entities from the source and injecting them into the summary while holding word count constant. The result is a fixed-length summary that grows progressively more specific with each iteration.

Unlike one-shot summarization, CoD forces entity accumulation across iterations, so the final summary contains more named specifics at the same length rather than generic paraphrase.

When to use it

  • Condensing verbose API documentation into a fixed-length reference
  • Compressing long requirements documents for executive review
  • Summarizing research papers while preserving key terminology
  • Reducing overlong agent skill files to meet length limits
  • Creating dense briefings from detailed incident reports

Example Workflow

Here's how your AI assistant might use this skill in practice.

INPUT

User asks: Summarize this 600-word architecture document in 80 words

AGENT
  1. 1Pass source text to cod-iteration agent as iteration 1 to produce a sparse base summary
  2. 2Parse the Missing_Entities line and Denser_Summary from the response
  3. 3Pass both the previous summary and original source to cod-iteration for iterations 2 through 5
  4. 4Track accumulated entities across all iterations
  5. 5Return the final dense summary and optionally the full iteration history
OUTPUT

An 80-word summary containing 9+ specific entities from the source, with optional YAML history showing entity accumulation per pass