Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
Many AI skills suffer from inconsistent output quality. This tool applies an autoresearch methodology to identify these failures, scoring outputs against binary evaluations and iteratively mutating prompts to eliminate errors.
It helps developers move beyond manual prompt engineering by automating the cycle of testing, scoring, and improving SKILL.md files through repeated execution loops.
WHO IT'S FOR
AI agent / automation builders
optimizing prompt performance via autonomous loops
Prompt engineers
reducing failure rates in complex AI skills
Technical leads / architects
benchmarking and validating AI skill reliability
Platform / DevEx teams
standardizing AI skill quality across repositories
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Runs per experiment, or acceptance of the default 5
Run interval, or acceptance of the default two minutes
Optional maximum number of experiment cycles
Make sure autoresearch is installed in your agent before trying this task.
Use the autoresearch skill to optimize the skill at <exact-path-to-target-SKILL.md>. Before running experiments, help me confirm 3–5 varied test prompts, 3–6 binary yes/no evaluation checks, the runs per experiment (default 5), the run interval (default every 2 minutes), and an optional experiment-cycle budget. Start with a baseline and a small capped run; report the proposed setup before making changes.
Claude