Terminal-Bench Loop
Drives one Terminal-Bench task through bounded smoke, diagnosis and board-gated fix cycles in Paperclip.
- Repository
- paperclipai/paperclip
- Stars
- 110,078
- Licence
- MIT
- Files
- 1 file
What it does
This skill teaches an AI agent how to run a repeatable, limited testing loop against a single Terminal-Bench task inside the Paperclip system. The agent organizes work as linked issues: a long-lived loop issue, one child per iteration with a smoke run, a diagnosis, and a fix proposal that needs explicit human (board) approval before any code change is made. It is an operational coordination playbook, not a code-writing skill; it says it does not authorize code changes by itself.
Great for
- Engineers on the Paperclip team iterating on Terminal-Bench smoke results
- Maintainers who need bounded, auditable diagnosis runs with approval gates before product fixes
Skip it if
- Building or changing the paperclip-bench harness itself, since the skill says to use normal engineering flow there
- Producing ranked or comparable benchmark submissions, since the skill produces non-comparable smoke runs by design
Try asking
- “Run terminal-bench/fix-git through a bounded Paperclip smoke loop until it passes”
- “Do the next iteration on our existing Terminal-Bench loop issue and diagnose the failure”
- “Iterate on Terminal-Bench until it passes, with board confirmation before any product fix”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the MIT licence and an attribution note.
terminal-bench-loop.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip terminal-bench-loop.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip terminal-bench-loop.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Set an iteration budget (typically 3 to 5) and a per-iteration wall-clock cap before starting, as the skill requires.
- Make sure the runner config enables a heartbeat-enabled agent; a bare Harbor command that creates an unassigned task is a setup failure, not a real diagnosis.
- Keep the benchmark command and artifact paths recorded on the loop issue so every iteration stays reproducible.
What's inside
- SKILL.md
Redistributed unmodified from paperclipai/paperclip under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.