autostar-web - Iteratively optimize measurable artifacts
Runs structured experiments to improve code, prompts, documents, configs, designs, or content through rubric-based evaluation and reflection.
Tags
Updated: 2026-10-02Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Run structured experiments
- Evaluate against rubrics
- Reflect at checkpoints
- Learn from attempts
- Manage optimization tracks
- Iterate until stopping criteria
Inputs
- Target artifact
- Optimization goal
- Evaluation rubric
- Hard constraints
- Soft constraints
- Verifier selections
- Experiment budget
- User approvals
- Memory connector
- Project memory pack
Outputs
- Improved artifact
- Experiment results
- Track scores
- Round reflections
- Final report
- Mission documents
Requirements
- Claude.ai custom Skills ZIP upload
- Web chat runtime
- User approval before execution
- Inline-evaluable verification checks
- No subprocess access
- Limited file read/write