summarizer

Generates recency-weighted session summaries from agent conversation transcripts via a subagent.

Updated Aug 23, 2026
One-click install
npx skills add https://github.com/fusuyfusuy/dot-agents --skill summarizer-fusuyfusuy
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: summarizer
Source: https://github.com/fusuyfusuy/dot-agents/tree/main/skills/summarizer
Command: npx skills add https://github.com/fusuyfusuy/dot-agents --skill summarizer-fusuyfusuy

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Long agent sessions accumulate large transcripts that are hard to review, and reading them directly pollutes the primary agent's context window. This Skill distills the active session into a compact, recency-weighted executive recap without consuming the orchestrating agent's working memory. ## Core Features & Use Cases - Subagent-Based Distillation: Delegates transcript parsing to an ephemeral subagent so the primary agent never reads raw logs directly. - 50/35/15 Recency Weighting: Allocates focus to the active frontier (latest turns), recent pivots, and foundational anchors in fixed proportions. - Standardized 4-Section Output: Returns a consistent markdown schema covering Active Frontier, Recent Trajectory, Foundational Anchors, and Next Action. - Use Case: After a long debugging session with many file edits and decisions, invoke the summarizer to get a scannable recap of uncommitted changes, pending decisions, and the single next step before handing off or resuming work. ## Quick Start Ask the agent to summarize the current session by typing /summarizer or requesting a session recap.

Frequently Asked Questions about summarizer

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

FAQPage Schema
How do I summarize an AI agent session transcript?▼

Invoke the summarizer explicitly with /summarizer or by requesting a session recap. It delegates transcript reading to an ephemeral subagent that applies recency weighting and returns a structured four-section markdown summary.

How does recency weighting work in session summarization?▼

The summary allocates roughly 50% focus to the latest one or two turns and in-flight state, 35% to recent pivots across turns N-1 to N-4, and 15% to foundational anchors like the original objective and early constraints.

Can the agent trigger session summarization automatically?▼

No. The skill sets disable-model-invocation to true, so it only runs when the user explicitly invokes it. This prevents autonomous background summarization from consuming context or interrupting active work.

Why does summarization use a subagent instead of the main agent?▼

Reading raw transcript logs directly would consume the primary agent's context budget and pollute its working memory. The subagent acts as an isolated pure function that parses the transcript and returns only the distilled markdown.

What are the limitations of transcript-based session summaries?▼

The summary depends on the completeness of the transcript log and is capped at 60 lines, so fine-grained detail from early turns may be compressed. It reflects stated requests versus verified on-disk state only as recorded in the transcript.