context-steward

Manages FlowDeck session context lifecycle with ingest, filter, prune, protect, summarize, persist, and telemetry.

20|3|Updated Apr 30, 2026
One-click install
npx skills add https://github.com/DVNghiem/FlowDeck --skill context-steward
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: context-steward
Source: https://github.com/DVNghiem/FlowDeck/tree/main/src/skills/context-steward
Command: npx skills add https://github.com/DVNghiem/FlowDeck --skill context-steward

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

FlowDeck sessions accumulate noise from tool outputs, rule loads, and multi-agent chatter, filling the context window and degrading response quality. Context Steward provides a unified lifecycle to keep context lean, relevant, and recoverable.

Core Features & Use Cases

  • Ingest: gather inputs from users, tools, and agent outputs
  • Filter & prune: remove duplicates and stale data while preserving critical context
  • Protect: guard user intent and safety records across phases
  • Telemetry: record pruning and context metrics for governance and improvement
  • Use Case: sustain long-running multi-agent sessions without memory bloat, while enabling safe rollbacks and auditing.

Quick Start

Load the Context Steward at session start and allow it to manage context balance across phases by applying ingest, filter, prune, protect, summarize, and persist steps during your FlowDeck session.

Frequently Asked Questions about context-steward

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

FAQPage Schema
How do I prevent context window limits from degrading multi-agent session quality?

Prevent context window limits by applying a deterministic context lifecycle that ingests, filters, prunes, and summarizes multi-agent session data. This removes duplicates and stale data while preserving critical context to keep responses high quality.

What is context pruning and how does it work during long-running agent workflows?

Context pruning is a deterministic pipeline that removes duplicates, purges errors, and compresses context during long-running agent workflows. It applies protected patterns to preserve essential information while keeping context within window limits.

How do I safely manage context lifecycle across plan, execute, and verify phases?

Safely manage context lifecycle across plan, execute, and verify phases by applying a unified pipeline of ingest, filter, prune, protect, summarize, and persist steps. This sustains long-running multi-agent sessions without memory bloat.

Can I audit context pruning and safety records in multi-agent workflows?

Yes, you can audit context pruning and safety records in multi-agent workflows using built-in telemetry. It records pruning and context metrics for governance, enabling safe rollbacks and continuous improvement of session management.

Does context management for FlowDeck sessions require any external dependencies?

Context management for FlowDeck sessions requires no external dependencies. It operates as a standalone advanced pipeline to orchestrate context lifecycle, ensuring safety and maintaining context balance natively.

When do I need protected patterns for context management in multi-agent sessions?

You need protected patterns for context management in multi-agent sessions when preserving user intent and safety records across phases is critical. They ensure essential context survives deterministic pruning and compression operations.