About autoresearch
A compatibility skill that directs research workflows to agi-mode's Hillclimb and behavioral comparisons to Eval. It reuses the user's existing goal, scope, and permissions rather than defining a separate experiment loop.
Key features
- 01Routes research to agi-mode Hillclimb
- 02Uses Eval for behavioral comparisons
- 03Reuses existing goals and permissions
- 04Supports recoverable snapshots without requiring Git
Use cases
- 01Run an optimization workflow through agi-mode
- 02Compare behavior using Eval
- 03Continue when Git is unavailable but recoverable snapshots exist
Install
Copy the command and run it in a terminal.
npx skills add https:/ / github.com/ xgent-ai/ skills --skill autoresearchAfter installing, type/autoresearchin Claude Code to confirm it loaded.
Details
From the project READMEAutoresearch
The maintained workflow now lives in agi-mode. Read and execute Hillclimb; it owns setup, baseline, safe recovery, experiment records, acceptance, plateaus and final confirmation. For behavioral comparisons it routes to Eval.
Reuse the user's existing goal, scope and permissions. This entry does not create a separate loop, mandate commits, or imply unlimited resources. Git is optional when equivalent recoverable snapshots are available.
Resolve the sibling path or the registered installation of agi-mode. If it is absent, explain that this compatibility entry requires agi-mode; do not invent a second implementation or claim to have run the missing workflow.