Chain of Density
PassAudited by ClawScan on May 1, 2026.
Overview
This is a purpose-aligned summarization skill with a simple local word-count helper, but it delegates iterations to a named subagent that users should trust before using it on sensitive text.
This skill appears safe for ordinary summarization tasks. Before using it on confidential material, confirm what `cod-iteration` subagent will run and whether your environment permits sharing the source text with that subagent.
Findings (2)
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.
The skill may fail if the subagent is absent, or it may use a separately installed subagent whose behavior was not reviewed with this skill.
The documented workflow relies on a named subagent, but the supplied manifest only includes SKILL.md and scripts/text_metrics.py, so that subagent's provenance and behavior are not visible in these artifacts.
Orchestrate 5 iterations via `cod-iteration` agent
Verify that the `cod-iteration` subagent is installed from a trusted source, or adapt the process to use a reviewed summarization prompt directly.
Sensitive documents summarized with this skill may be exposed to the subagent/model context used for each iteration.
The skill intentionally sends the original source text through repeated subagent calls, which is purpose-aligned for densification but creates an inter-agent data flow users should be aware of.
Pass SOURCE text in every iteration for entity discovery
Use this skill only with text you are permitted to share with the configured agent/subagent, and avoid including secrets unless the execution environment is trusted.