summarize

Compress documents, issues, code, and sessions into faithful evidence-grounded summaries.

Updated Sep 10, 2026
One-click install
npx skills add https://github.com/sergeiwallace/skills --skill summarize-sergeiwallace
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: summarize
Source: https://github.com/sergeiwallace/skills/tree/main/skills/summarize
Command: npx skills add https://github.com/sergeiwallace/skills --skill summarize-sergeiwallace

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Condensing long documents, issue threads, codebases, or session logs into a trustworthy brief is slow and error-prone, and naive summaries often drop decisions, caveats, or contradictions. This Skill produces faithful, reader-targeted summaries that preserve requirements, chronology, and uncertainty without adding outside facts. ## Core Features & Use Cases - Typed route adapters: Automatically routes sources to document, issue, code, topic, or session adapters that retain the fields that matter for each type, such as causal chains for bugs or delivery contracts for design docs. - Evidence integrity controls: Builds a source manifest tracking read, partial, missing, or denied access, labels inference separately from source-stated fact, and emits evidence pointers for code reconstructions and disputed claims. - Configurable depth and format: Supports brief, standard, and deep modes with briefing, bullets, or prose formats, plus must-keep items and a declared reader and purpose. - Use Case: After a long design review, ask for a standard-depth summary of the design doc and linked issue thread for a teammate deciding next steps, and receive a structured brief with decisions, risks, open questions, and evidence pointers. ## Quick Start Ask the assistant to summarize the design document at docs/proposal.md in standard mode for a technical collaborator deciding what to do next.

Frequently Asked Questions about summarize

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I summarize a design document or issue thread with an AI agent?

Invoke the summarize skill with the source paths or issue references, optionally specifying reader, purpose, mode, and format. It routes the input to a document or issue adapter, retains decisions and chronology, and renders a structured brief with risks and next actions.

How do I control summary length and format?

Use the mode flag with brief, standard, or deep values and the format flag with briefing, bullets, or prose. Brief yields roughly 5-8 bullets, standard 400-800 words, and deep 1,000-2,000 words with evidence pointers.

What is the difference between summarize and summary skills?

The summarize skill compresses caller-selected evidence like documents, issues, and code. The summary skill is a fixed end-of-batch session retrospective; summarize delegates to it only for explicit completed or paused session requests.

Can summarize handle sources it cannot fully read?

Yes. It records each source in a manifest with access marked read, partial, missing, or denied, discloses the gaps, and never implies unread material was covered. If the readable subset cannot answer the purpose, it reports the blocker.

Does the summary distinguish facts from inference?

Yes. Every claim must be entailed by readable evidence or explicitly labeled as inference. Code summaries separate observed behavior, source-stated rationale, and inference into distinct claim classes.