Latent Briefing
Skill ActiveThis skill should be used when the user asks to "share memory between agents", "KV cache compaction for multi-agent", "orchestrator worker context", "latent briefing", "reduce worker tokens", "cross-agent memory without summarization", or discusses Attention Matching compaction, recursive language models with workers, or token explosion in hierarchical agents.
To provide a detailed explanation and implementation guidance for using Latent Briefing to optimize memory sharing and reduce token costs in hierarchical multi-agent systems.
Features
- Explains representation-level memory sharing via KV cache compaction.
- Details the adaptation of Attention Matching for multi-agent inference.
- Discusses three inference-time modifications: task-guided queries, shared masks, and MAD thresholding.
- Outlines infrastructure preconditions and decision frameworks for choosing memory sharing mechanisms.
Use Cases
- Designing orchestrator-worker systems needing to share prior state efficiently.
- Evaluating KV cache compaction as an alternative to text summarization for cross-agent state transfer.
- Debugging token explosion in recursive or hierarchical agent graphs.
- Implementing or studying task-conditioned selective retention in LLM inference.
Non-Goals
- Replacing text-based summarization or RAG where those methods are sufficient or preferable.
- Providing a deployable tool; this skill is for conceptual understanding and implementation guidance.
- Working with API-only stacks where KV state is inaccessible.
Practical Utility
- info:Production readinessThe skill outlines a detailed technical approach and use cases, but its readiness for production hinges on the user's ability to control the worker inference runtime for KV state manipulation.
Maintenance
- warning:Commit recencyThe last commit was on 2026-04-14, over 3 months ago, suggesting potential unmaintained status.
Trust
- warning:Issues Attention6 issues opened and 2 closed in the last 90 days, indicating a low closure rate (33%) and potentially slow maintainer response.
Installation
First, add the marketplace
/plugin marketplace add muratcankoylan/Agent-Skills-for-Context-Engineering/plugin install Agent-Skills-for-Context-Engineering@context-engineering-marketplaceQuality Score
Trust Signals
Similar Extensions
Context Optimization
100This skill should be used when the user asks to "optimize context", "reduce token costs", "improve context efficiency", "implement KV-cache optimization", "partition context", or mentions context limits, observation masking, context budgeting, or extending effective context capacity.
Stream Chain
99Stream-JSON chaining for multi-agent pipelines, data transformation, and sequential workflows
Context Compression
100This skill should be used when the user asks to "compress context", "summarize conversation history", "implement compaction", "reduce token usage", or mentions context compression, structured summarization, tokens-per-task optimization, or long-running agent sessions exceeding context limits.
Init
100Creates, updates, or optimizes an AGENTS.md file for a repository with minimal, high-signal instructions covering non-discoverable coding conventions, tooling quirks, workflow preferences, and project-specific rules that agents cannot infer from reading the codebase. Use when setting up agent instructions or Claude configuration for a new repository, when an existing AGENTS.md is too long, generic, or stale, when agents repeatedly make avoidable mistakes, or when repository workflows have changed and the agent configuration needs pruning. Applies a discoverability filter—omitting anything Claude can learn from README, code, config, or directory structure—and a quality gate to verify each line remains accurate and operationally significant.
External Context
100Invoke parallel document-specialist agents for external web searches and documentation lookup
Swarm Orchestration
100Orchestrate multi-agent swarms with agentic-flow for parallel task execution, dynamic topology, and intelligent coordination. Use when scaling beyond single agents, implementing complex workflows, or building distributed AI systems.