A collection of Claude Code skills for enhanced AI-assisted development workflows.
| Skill | Description |
|---|---|
| scope-check | Context window evaluation and chunking strategy tool. Prevents mid-implementation context collapse by validating document size before writing large design documents or implementation plans. |
/plugin marketplace add dvdarkin/claude-skills
/plugin install scope-check@dvdarkinRestart Claude Code to pick up the new skill.
Clone the repository and copy desired skills to your Claude skills directory:
git clone https://github.com/dvdarkin/claude-skills.git
cp -r claude-skills/skills/scope-check ~/.claude/skills/Evaluates specification documents against Claude Code's context window constraints. Provides grounded estimations of whether a spec will fit within safe context bounds and recommends chunking strategies when it won't.
Features:
- Automatic activation before writing design documents or implementation plans
- Traffic light system (GREEN/YELLOW/ORANGE/RED) for context budget status
- Interactive chunking recommendations for large specs
- Research-backed expansion estimates based on language and TDD level
Quick Start:
# Basic check
/scope-check docs/feature-spec.md
# With explicit overrides
/scope-check docs/feature-spec.md --stack typescript --tdd comprehensiveExample Output: See scope-check-example.md for a real-world example analyzing a complex AI Virtual Car Showroom spec.
Research Foundation:
- QSM Database: Function Point to LOC ratios (2,192 projects)
- IFPUG ISO/IEC 20926:2009: Function Point counting standards
- Liu et al. (2024): "Lost in the Middle" - context position bias
- Chroma Research: "Context Rot" study across 18 LLMs
- RULER Benchmark: Long-context model evaluation
See scope-check documentation for full details.
Contributions welcome. If you have a skill that could benefit the Claude Code community, feel free to submit a PR.
Issues and PRs: https://github.com/dvdarkin/claude-skills
MIT License - see LICENSE for details.
Denis Darkin (dvdarkin@gmail.com)