Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →clawbio/clawpathy-autoresearch
Eval-driven skill tuning. Given a task and an LLM-judge rubric, iteratively rewrites a SKILL.md until a downstream executor agent performs well against the judge. Low-code: all evaluation is LLM-as-judge, not deterministic Python.
Passed 17 of 21 test cases with the skill, and 11 of 21 without it.
and 5 more in the full report
6 fixed13 passed either way5 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 6 test cases the skill fixed. Same model, same prompt, with and without the skill.
On iteration 4, the score was 15. On iteration 5, the skill modification resulted in a score of 18 (where lower score is better). Should we update our best skill snapshot to iteration 5 because it generated interesting new code?
| Checked | Without skill | With skill |
|---|---|---|
| The response specifies reverting the skill modification from iteration 5. | Missed | Passed |
| The response states that lower numeric scores represent superior performance. | Passed | Passed |
The instructions your model reads when the skill loads.
Eval-driven skill development. The system iteratively rewrites a SKILL.md
so a downstream executor agent performs better at a task class, as judged
by an LLM against a paper/task-specific rubric.
propose (sonnet) → execute (sonnet, shell) → judge (opus, rubric)
↑ │
└──────── feedback: verdict + recommended edits ────────┘
early_stop_n consecutive regressions.You (the agent reading this) don't run the loop yourself. You dispatch subagents to build the workspace, then hand off to the Python loop.
Dispatch a subagent with prompts/scout.md to research the paper/task.
Report key findings to the user in a few lines.
Have a conversation. Ask ONE question at a time, multiple-choice where helpful. Agree on:
test cases passed, vs 11 / 21 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.