Context Engineering Guide
Teaches an agent how to manage what fills its context window: what to load, prune, or hand off.
- Repository
- mvschwarz/openrig
- Stars
- 8,553
- Licence
- Apache-2.0
- Files
- 1 file
What it does
When installed, this skill gives an AI agent a reference framework for deciding what information goes into its working memory (the context window) at each step. It explains why agents drift or forget constraints, how to keep always-loaded files lean, when to use subagents versus a single agent, and how to plan handoffs and summarization. It is a 2024–2025 research snapshot that is explicitly marked as provisional and not authoritative for current models.
Great for
- Developers designing or debugging a multi-step AI agent setup that drifts or repeats itself
- Teams deciding what should load at startup versus on demand for an agent
- Students learning how context windows and attention limits affect agent behavior
Skip it if
- Rewording prompt tone or style, which the skill says it does not cover
- Choosing which model to pin, since the skill excludes model selection
- Relying on its specific numeric context ceilings as current facts, because the file itself flags them as perishable and in need of re-verification
Try asking
- “Our agent keeps forgetting constraints halfway through a long task. Help me find what in its context is missing, stale, or bloated.”
- “Should this job use one agent or several subagents? Walk me through the tradeoffs.”
- “Review my always-loaded instructions file and tell me how to shrink it to only high-signal content.”
- “Design a handoff so a second agent can pick up where the first one stopped.”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.
context-engineering.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip context-engineering.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip context-engineering.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Treat the numeric thresholds and vendor claims as historical; ask the agent to re-verify them against current model documentation before acting.
- Give the skill a concrete problem (a specific agent that misbehaves) so it applies the checklist instead of giving general theory.
- Check its guidance on compaction and multi-agent design against your own measured results, since the file says measured practice takes precedence.
What's inside
- SKILL.md
Redistributed unmodified from mvschwarz/openrig under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.