context-management

Keeps working context focused through targeted reads, compact handoffs, and bounded delegation.

2|Updated May 14, 2026
One-click install
npx skills add https://github.com/AnthonyPoschen/agent-skills --skill context-management-anthonyposchen
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: context-management
Source: https://github.com/AnthonyPoschen/agent-skills/tree/main/skills/context-management
Command: npx skills add https://github.com/AnthonyPoschen/agent-skills --skill context-management-anthonyposchen

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Large logs, long files, broad search results, and multi-phase investigations quickly crowd an agent's working context, degrading decision quality. This Skill provides guardrails for reading only what matters, preserving compact records, and delegating bounded investigations so the context stays focused on the decisions that remain. ## Core Features & Use Cases - Question-Driven Reading: State the question a file or log must answer before opening it, then read only the relevant section, range, or filter instead of the whole source. - Compact Working Records: Preserve only goals, decisions, evidence locations, open questions, and next actions between phases instead of pasting verbose transcripts. - Bounded Delegation: Delegate large independent investigations with a clear scope and required facts, while keeping small coupled reads in the main thread. - Use Case: While debugging a production incident across gigabytes of logs and dozens of source files, use this Skill to extract only the relevant log windows, maintain a short evidence note, and delegate a bounded trace analysis without flooding the context. ## Quick Start Use the context-management skill to investigate this large log directory and report only the evidence needed to decide the next fix.

Frequently Asked Questions about context-management

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

FAQPage Schema
How to read large log files efficiently with an AI agent?▼

State the question the log must answer first, then use targeted searches, ranges, or filters rather than reading the entire file. Record the fact, its evidence location, and the next action instead of rereading.

When should I delegate an investigation to a subagent?▼

Delegate when the investigation has a clear question, can be explored independently, and the main task can proceed from a concise evidence-backed summary. Keep small, tightly coupled reads in the main thread.

What are the limitations of context window management techniques?▼

Summarization can lose exact wording needed later, and delegation adds overhead for small coupled reads. Inspect important evidence yourself before consequential decisions and avoid creating agents merely to skip a few file reads.

Does context management work for multi-phase coding tasks?▼

Yes. Between phases, preserve only the goal, constraints, decisions made, direct evidence with locations, open questions, and the next action. Make a compact handoff record before starting a broad new phase or large fan-out.