Reusable Claude Code skills for software development workflows. Works with any project, any language.
# In Claude Code
/plugin marketplace add 8-bit-sheep/code-harness-skills/plugin install skill-builder@8-bit-sheep/code-harness-skills
/plugin install sprint-planner@8-bit-sheep/code-harness-skills
/plugin install release-manager@8-bit-sheep/code-harness-skills
# ... etcgit clone https://github.com/8-bit-sheep/code-harness-skills
# In Claude Code:
/plugin marketplace add ./code-harness-skills| Skill | Description |
|---|---|
| skill-builder | Create, test, evaluate, and optimize Claude Code skills with eval-driven iteration |
| headless-runner | Run Claude Code programmatically in headless mode for CI/CD, automation, and agent workflows |
| cloud-setup | Set up development environments in cloud/mobile Claude Code sessions |
| Skill | Description |
|---|---|
| codebase-organizer | Monitor file sizes and split large files into AI-friendly modules |
| test-coverage-guardian | Analyze test coverage, identify gaps, detect dead code, enforce quality gates |
| perf-reviewer | Review code for performance issues, run benchmarks, identify optimizations |
| Skill | Description |
|---|---|
| sprint-planner | Analyze design docs, calculate velocity, create realistic sprint plans |
| sprint-executor | Execute sprint plans with TDD, milestone tracking, and checkpoint validation |
| design-doc-creator | Create structured design documents with systemic analysis |
| release-manager | Automate releases with version bumps, changelogs, tags, and CI verification |
| Skill | Description |
|---|---|
| github-issue-triage | Monitor and triage GitHub issues against design docs and implementation status |
| agent-inbox | Cross-agent messaging with inbox management and handoff patterns |
Skills use environment variables with sensible defaults. Each skill documents its configuration in a ## Configuration section within its SKILL.md.
Common variables:
| Variable | Default | Used By |
|---|---|---|
TEST_CMD |
make test |
Most skills |
LINT_CMD |
make lint |
Most skills |
BUILD_CMD |
make build |
Most skills |
SOURCE_DIR |
. |
codebase-organizer |
DESIGN_DOCS_DIR |
design_docs/ |
design-doc-creator, sprint-planner, github-issue-triage |
COVERAGE_TARGET |
80 |
test-coverage-guardian |
SPRINT_STATE_DIR |
.sprint-state |
sprint-planner, sprint-executor |
The skill-builder skill is a meta-skill that helps you create, test, and optimize new skills:
User: "Create a skill for database migration management"
Claude: [Uses skill-builder to scaffold, test, and iterate on a new skill]
User: "Check file sizes and split any large files"
Claude: [Uses codebase-organizer to find files >800 lines and propose splits]
User: "Plan a sprint from the auth-redesign design doc"
Claude: [Uses sprint-planner to analyze velocity and create a milestone-based plan]
User: "Execute the sprint"
Claude: [Uses sprint-executor with TDD, running tests at each checkpoint]
Each skill follows the progressive disclosure pattern:
skill-name/
SKILL.md # Always loaded (core docs, <300 lines)
scripts/ # Execute as needed (automation)
resources/ # Load on demand (detailed references)
See the skill-builder for the complete guide to creating, testing, and optimizing skills.
Key principles:
- Pushy descriptions: List specific trigger scenarios to combat under-triggering
- Progressive disclosure: Keep SKILL.md lean, details in resources/
- Eval-driven: Test with the skill-builder's eval framework before shipping
- Fork this repository
- Create a new skill in
skills/your-skill-name/ - Include a
SKILL.mdwith YAML frontmatter (name+description) - Add
scripts/andresources/as needed - Run
skill-builder/scripts/validate_skill.sh skills/your-skill-name/ - Submit a PR
MIT