context-curation
Analyzes staged git changes and evaluates agentic context files (Claude, Codex, Cursor, etc.) to suggest additions or removals. Use after staging changes before committing.
Best use case
context-curation is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Analyzes staged git changes and evaluates agentic context files (Claude, Codex, Cursor, etc.) to suggest additions or removals. Use after staging changes before committing.
Teams using context-curation 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-curation/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How context-curation Compares
| Feature / Agent | context-curation | 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?
Analyzes staged git changes and evaluates agentic context files (Claude, Codex, Cursor, etc.) to suggest additions or removals. Use after staging changes before committing.
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 Curation Analyze staged git changes and evaluate whether agentic context files need to be updated to reflect those changes. ## When to Use This Skill - After staging changes, before committing - When introducing new patterns, conventions, or APIs - When deprecating or removing significant functionality - When refactoring code that may invalidate existing context documentation - Periodically to audit context file freshness ## Supported Context File Locations This skill scans for agentic context files across multiple platforms: | Platform | Locations | |----------|-----------| | Claude Code | `.claude/`, `CLAUDE.md` | | Codex | `.codex/`, `codex.md` | | Cursor | `.cursor/`, `.cursorrules` | | Aider | `.aider/`, `.aider.conf.yml` | | GitHub Copilot | `COPILOT.md`, `.github/copilot-instructions.md` | | Generic | `AGENTS.md`, `AI.md`, `CONTEXT.md` | | Project docs | `docs/`, `README.md` (architecture sections) | ## Analysis Workflow ### Step 1: Inspect Staged Changes Run `git diff --staged` to understand what's being committed: - New files and their purposes - Modified functions, classes, or modules - Deleted code and removed dependencies - Changed APIs, interfaces, or contracts ### Step 2: Scan Existing Context Files Locate all agentic context files in the repository and extract: - Documented patterns and conventions - Architecture decisions and rationale - API references and usage examples - Project-specific rules and guidelines ### Step 3: Compare and Analyze Cross-reference staged changes against existing context: - **New patterns**: Code introduces conventions not yet documented - **API changes**: Public interfaces modified or extended - **Deprecated code**: Removed functionality still referenced in context - **Stale references**: Context mentions files, functions, or patterns that no longer exist ### Step 4: Generate Recommendations Produce actionable suggestions with clear rationale. ## Output Format ### Additions Needed For each suggested addition: ``` ## ADD: [Brief description] **Context file**: [path to file that should be updated] **Rationale**: [Why this should be documented] **Suggested content**: [Proposed text to add] **Evidence from staged changes**: - [file:line] - [relevant code snippet or change] ``` ### Removals Needed For each suggested removal: ``` ## REMOVE: [Brief description] **Context file**: [path to file containing outdated content] **Current content**: [text that should be removed or updated] **Rationale**: [Why this is now stale] **Evidence from staged changes**: - [file] - [deleted or changed code that invalidates this] ``` ### No Changes Needed If context files are already aligned with staged changes: ``` ## Context files are up to date Reviewed [N] context files against staged changes. No updates recommended. ``` ## Instructions 1. First, run `git diff --staged --stat` to get an overview of staged changes 2. Run `git diff --staged` to examine the actual changes in detail 3. Search for all supported context file locations listed above 4. Read each context file found 5. Analyze whether staged changes introduce patterns, APIs, or conventions that should be documented 6. Analyze whether staged changes remove or deprecate anything referenced in context files 7. Present findings using the output format above 8. If no context files exist, suggest creating one appropriate for the project
Related Skills
extracting-ai-context
Extracts and manages AI context (skills, AGENTS.md) from workflow-kotlin library JARs. Use when setting up AI tooling for a workflow-kotlin project, updating skills after a library version change, or configuring agent-specific directories.
create-agent-with-sanity-context
Build AI agents with structured access to Sanity content via Context MCP. Covers Studio setup, agent implementation, and advanced patterns like client-side tools and custom rendering.
context-optimizer
Analyzes Copilot Chat debug logs, agent definitions, skills, and instruction files to audit context window utilization. Provides log parsing, turn-cost profiling, redundancy detection, hand-off gap analysis, and optimization recommendations. Use when optimizing agent context efficiency, identifying where to add subagent hand-offs, or reducing token waste across agent systems.
context-fundamentals
Understand the components, mechanics, and constraints of context in agent systems. Use when designing agent architectures, debugging context-related failures, or optimizing context usage.
context-engineering
Use when designing agent system prompts, optimizing RAG retrieval, or when context is too expensive or slow. Reduces tokens while maintaining quality through strategic positioning and attention-aware design.
context-degradation
Recognize patterns of context failure: lost-in-middle, poisoning, distraction, and clash
context-assembler
Assembles relevant context for agent spawns with prioritized ranking. Ranks packages by relevance, enforces token budgets with graduated zones, captures error patterns for learning, and supports configurable per-agent retrieval limits.
Codebase context
Create a lightweight codebase_context.md that anchors the idea in the existing repo (modules, constraints, extension points). Generic framework prompt.
agent-context-system
A persistent local-only memory system for AI coding agents. Two files, one idea — AGENTS.md (committed, shared) + .agents.local.md (gitignored, personal). Agents read both at session start, update the scratchpad at session end, and promote stable patterns over time. Works across Claude Code, Cursor, Copilot, Windsurf. Subagent-ready. No plugins, no infrastructure, no background processes.
add-route-context
为Flutter页面添加路由上下文记录功能,支持日期等参数的AI上下文识别。当需要让AI助手通过"询问当前上下文"功能获取页面状态(如日期、ID等参数)时使用。适用场景:(1) 日期驱动的页面(日记、活动、日历等),(2) ID驱动的页面(用户详情、订单详情等),(3) 任何需要AI理解当前页面参数的场景
localsetup-context
Localsetup v2 framework context - overview, invariants, and skills index. Load first when working in a repo that uses Localsetup v2. Use when starting work in this repo or when user asks about framework rules.
context7-skills
Use when a user asks to search, install, list, or remove skills with the Context7 ctx7 skills CLI (including npx ctx7) and needs correct subcommands or client flags.