context-engineer
Determines WHAT context an agent needs and packages it optimally. Actions: analyze (identify needed context), load (assemble from sources), prune (trim to token budget), inject (write to .claude/context-payload.md). Use when: (1) before spawning teammates, (2) context window is limited, (3) multi-source context assembly. Triggers: /context-engineer, 'prepare context', 'package context', 'context for agent'.
Best use case
context-engineer is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Determines WHAT context an agent needs and packages it optimally. Actions: analyze (identify needed context), load (assemble from sources), prune (trim to token budget), inject (write to .claude/context-payload.md). Use when: (1) before spawning teammates, (2) context window is limited, (3) multi-source context assembly. Triggers: /context-engineer, 'prepare context', 'package context', 'context for agent'.
Teams using context-engineer should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/context-engineer/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How context-engineer Compares
| Feature / Agent | context-engineer | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Determines WHAT context an agent needs and packages it optimally. Actions: analyze (identify needed context), load (assemble from sources), prune (trim to token budget), inject (write to .claude/context-payload.md). Use when: (1) before spawning teammates, (2) context window is limited, (3) multi-source context assembly. Triggers: /context-engineer, 'prepare context', 'package context', 'context for agent'.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Context Engineer v3.0 Determine what context an agent needs and package it within token budget. ## Actions ### `/context-engineer analyze <task>` Identify what context sources are relevant: 1. Parse task description for domain keywords 2. Check for DESIGN.md (frontend tasks) 3. Check for .spec.md files (spec-driven tasks) 4. Check memory sources (vault, handoffs, ledgers) 5. Check codebase files (Glob/Grep for related code) 6. Output: ranked list of context sources with estimated tokens ### `/context-engineer load <task>` Assemble context from identified sources: 1. Run `analyze` if not already done 2. Read each source, extract relevant sections 3. Prioritize by relevance score 4. Output: assembled context block ### `/context-engineer prune <budget>` Trim assembled context to fit token budget: 1. Default budget: 8000 tokens 2. Remove low-relevance sections first 3. Summarize long sections instead of including full text 4. Preserve: interfaces, invariants, design tokens (high signal) 5. Output: pruned context within budget ### `/context-engineer inject` Write packaged context to `.claude/context-payload.md`: 1. Write pruned context to file 2. The `ralph-subagent-start.sh` hook reads this file and injects into teammate prompt 3. File is gitignored (per Item 0a) ## Context Sources (Priority Order) | Source | When Relevant | Token Cost | |---|---|---| | .spec.md | Task has a spec | ~500-2000 | | DESIGN.md | Frontend task | ~1000-3000 | | Related code files | Always | ~500-5000 | | Memory (vault) | Prior work exists | ~200-500 | | Handoffs | Cross-session | ~300-800 | | Vault (future) | Curated knowledge | ~200-1000 | ## Integration - **Orchestrator Step 5 (DELEGATE)**: invoke before spawning teammate - **ralph-subagent-start.sh**: reads `.claude/context-payload.md` if exists - **Token budget**: default 8000, adjustable per task complexity ## Anti-Rationalization | Excuse | Rebuttal | |---|---| | "The agent has enough context from the prompt" | Agents lose 40% accuracy without structured context. Package it. | | "Context engineering is overhead" | 5 seconds of packaging saves 5 minutes of wrong output. | | "I'll just include everything" | Token flooding degrades quality. Prune to signal. |
Related Skills
senior-software-engineer
Global skill enforcing senior software engineering best practices
sec-context-depth
Comprehensive AI code security review using 27 sec-context anti-patterns. Use for code review when security vulnerabilities are suspected, especially for AI-generated code.
context7-usage
Patterns for using Context7 MCP for library documentation (v2.25)
worktree-pr
Manage git worktrees with PR workflow and multi-agent review (Claude + Codex). Use when developing features in isolation with easy rollback.
vercel-react-best-practices
React and Next.js performance optimization guidelines from Vercel Engineering. Use when writing, reviewing, or refactoring React/Next.js code. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.
vault
Living knowledge base management. Actions: search (query vault), save (store learning), index (update indices), compile (raw->wiki->rules graduation), init (create vault structure). Follows Karpathy pipeline: ingest->compile->query. Use when: (1) searching accumulated knowledge, (2) saving learnings, (3) compiling raw notes into wiki, (4) initializing a new vault. Triggers: /vault, 'vault search', 'knowledge base', 'save learning'.
testing-anti-patterns
Custom skill for testing-anti-patterns
task-visualizer
Visualize task dependencies and progress (Gastown-style)
task-classifier
Classifies task complexity (1-10) for model and agent routing
task-batch
Autonomous batch task execution with PRD parsing, task decomposition, and continuous execution until all tasks complete. Uses /orchestrator internally. Stops only for major failures (no internet, token limit, system crash). Use when: (1) processing task lists autonomously, (2) PRD-driven development, (3) batch feature implementation. Triggers: /task-batch, 'batch tasks', 'process PRD', 'run task queue'.
tap-explorer
Tree of Attacks with Pruning for systematic code analysis
stop-slop
A skill for removing AI-generated writing patterns ('slop') from prose. Eliminates telltale signs of AI writing like filler phrases, excessive hedging, overly formal language, and mechanical sentence structures. Use when: writing content that should sound human and natural, editing AI-generated drafts, cleaning up prose for publication, or any content that needs to sound authentic rather than AI-generated. Triggers: 'stop-slop', 'remove AI tells', 'clean up prose', 'make it sound human', 'edit AI writing'.