Context Packer

Distill messy chat logs and project notes into dense, action-oriented briefs optimized for AI agents.

Install
cmdop skills install agensi-context-packer

Context Packer is a skill that acts as a pre-processor for AI agents. It takes large, unstructured inputs — chat logs, Slack threads, project notes — and produces compact briefs that prioritize information density over readability. The goal is to retain structural details such as specific constraints, exact file paths, and hard numbers while discarding conversational filler, redundant reasoning, and social niceties that consume context window space without contributing to agent task completion.

This matters in multi-agent pipelines and long-running sprints where context window exhaustion degrades output quality. By stripping out low-value content before passing context to a downstream agent, Context Packer reduces token usage on follow-up prompts and keeps agents focused on the actual task rather than tangential details present in raw logs.

The skill also explicitly surfaces blockers and open questions, which helps prevent stalled workflows when an agent or developer picks up mid-project. It is designed to create consistent, repeatable handoffs between agents or between stages of a development sprint.

Context Packer is not a general-purpose summarizer. It does not optimize output for human readability and does not produce narrative summaries. It is the wrong choice when the goal is a polished human-facing document or when the source material is already well-structured and compact.

Use cases

  • Compress a long Slack thread into a structured brief before passing it to a coding agent
  • Extract specific file paths, constraints, and blockers from disorganized project notes
  • Create consistent context handoffs between agents at different pipeline stages
  • Reduce token costs by stripping filler before feeding logs into follow-up prompts
  • Surface open questions and blockers from meeting transcripts for a planning agent
  • Pre-process sprint retrospectives into action-oriented summaries for downstream agents

When to use it

  • When source context is verbose and unstructured, such as raw chat logs or meeting notes
  • When operating in a multi-agent pipeline that requires consistent handoff format
  • When context window limits are a concern and token reduction is a priority
  • When downstream agents are drifting due to noisy or tangential context

When not to use it

  • When the output needs to be readable and polished for a human audience
  • When source material is already concise and well-structured
  • When a narrative summary rather than a dense brief is required
  • When no tools or transport details are provided and integration path is unclear