How dreambigou/eli5 works
ELI5 is a Claude Code skill that adapts explanations to an intended audience. Its principal workflow is described in the README, but the skill instructions themselves were not sampled, so the runtime stages are represented at a documented, high level. The repository also includes an evaluation workflow; its runner excerpt shows calls to the Claude CLI, while omitted code limits detail about orchestration.
Main entry points: run-evals.py Eval runner · SKILL.md ELI5 skill · evals.json Eval cases
Components 10
Explanation skill
- ELI5 skill
skills/eli5/SKILL.md - Audience detection
skills/eli5/SKILL.md - Style calibration
skills/eli5/SKILL.md - Adapted explanation
Evaluation workflow
- Eval runner
eli5-workspace/run-evals.py - Eval cases
eli5-workspace/evals.json - Evaluation outputs
eli5-workspace/run-evals.py
Other components
- User
- Claude Code
- Claude CLI
Connections 10
- User → Claude Code: submits prompt
- Claude Code → ELI5 skill: uses skill
- ELI5 skill → Audience detection: guides detection
- Audience detection → Style calibration: informs style
- Style calibration → Adapted explanation: shapes response
- Claude Code → Adapted explanation: returns response
- Eval runner → Eval cases: loads cases
- Eval runner → Claude CLI: runs prompts
- Eval runner → Claude CLI: grades responses
- Eval runner → Evaluation outputs: writes outputs