terminal-bench-loop

A bounded workflow for running one Terminal-Bench task through repeated testing, diagnosis, and approved fixes. Terminal-Bench is a benchmark for evaluating coding agents on terminal-based tasks.

In plain words
What is it for?
Use it to run a Terminal-Bench smoke loop, inspect failures, coordinate approved implementation fixes, and continue until the task passes or is clearly blocked.
Why use it?
It keeps repeated benchmark attempts organized and prevents fixes from being made without the required issue tracking and approval. It also records whether the task passed or has a specific blocker.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/paperclipai/paperclip/terminal-bench-loop
Any agent
npx skills add paperclipai/paperclip --skill terminal-bench-loop
Clone the repo
git clone --depth 1 https://github.com/paperclipai/paperclip

Made for: Claude Code, Codex.

Per session 54 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 5,451 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00054 $0.05451
Opus 5 $0.00027 $0.02726
Sonnet 5 $0.00011 $0.01090
Haiku 4.5 $0.00005 $0.00545

Measured yesterday against content hash 4acbe1f61931, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

terminal-bench-loop scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.agents/skills/terminal-bench-loop/SKILL.md · 229 lines

How it starts

The opening of the file, as written. The whole thing — 229 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Terminal-Bench Loop

A repeatable operating skill for driving one Terminal-Bench problem to a passing smoke through Paperclip, with explicit issue topology, bounded runs, board-gated product fixes, and worktree continuity.

This skill is operational + diagnostic, not engineering. It coordinates issues, artifacts, and approvals around a Terminal-Bench loop. It does not authorize code changes — every accepted product fix lands as a separate implementation child issue after a board confirmation.

Canonical execution model: read doc/execution-semantics.md before starting a loop or moving any loop issue. Every loop issue must rest in a state the doc allows: terminal (done/cancelled), explicitly live (active run / queued wake), explicitly waiting (in_review with participant/interaction/approval), or explicit recovery/blocker (blocked with blockedByIssueIds and a named owner).

When to use

Trigger on an assignment whose title or body matches any of:

  • "run Terminal-Bench in a loop", "loop <task-name> through Paperclip"
  • "drive Terminal-Bench fix-git", "iterate on Terminal-Bench until it passes"
  • "Terminal-Bench smoke loop", "bench loop", "smoke loop on <task-name>"
  • An attached link to a Terminal-Bench loop parent issue, plus a request to do another iteration

Also use when the user hands you an existing top-level loop issue and asks for the next iteration, diagnosis, or rerun.

When NOT to use

  • The assignment is to build or change paperclip-bench itself (Harbor adapter, wrapper, telemetry). Use normal engineering flow on that repo.
  • The assignment is to submit a benchmark result for ranking. This skill produces smoke/non-comparable runs by design — escalate full-suite or comparable runs to BenchmarkQualityManager.
  • The assignment is a normal Paperclip product bug not surfaced by a Terminal-Bench loop. Use normal investigation.
  • You have not been granted permission to install or assign company skills, and the asker actually wants library mutation. Hand that step to an authorized skill-library owner.

Read the full file on GitHub · 229 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 229 lines · 54 tokens per session scan A 4acbe1f61931

Subscribe to this mod's changes

terminal-bench-loop is a skill published in the GitHub repository paperclipai/paperclip (79,688 stars, last pushed yesterday), licensed MIT. It adds 54 tokens to every session and 5,451 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

systematic-debugging

Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes.

obra/superpowers · 21 tokens

next-cache-components-adoption

Turn on Cache Components in a Next.js app and resolve the blocking routes it surfaces. Use when the user wants to enable, adopt, or migrate to Cache Components, flip the cacheComponents flag, work through a flood of blocking-prerender / instant validation errors, run the cache-components-instant-false codemod, or…

vercel/next.js · 95 tokens

babysit-pr

Babysit a GitHub pull request after creation by continuously polling review comments, CI checks/workflow runs, and mergeability state until the PR is merged/closed or user help is required. Diagnose failures, retry likely flaky failures up to 3 times, auto-fix/push branch-related issues when appropriate, and keep…

openai/codex · 114 tokens

imagegen

Generate or edit raster images when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, or transparent-background cutouts. Use when Codex should create a brand-new image, transform an existing image, or derive visual variants from references, and the output…

openai/codex · 113 tokens

cpu-profile-analysis

Analyze V8/Chrome CPU profiles (.cpuprofile) and DevTools trace files (Trace-.json). Use when: profiling performance, investigating slow functions, comparing code paths, finding bottlenecks, analyzing timeToRequest, understanding call trees from sampling profiler data, analyzing layout/paint/rendering, investigating…

microsoft/vscode · 71 tokens

next-cache-components-optimizer

Drive a Next.js route to instant navigation by setting up an agentic loop, under Cache Components / PPR, on initial load (hard navigation) and client-side navigation (soft navigation). Encode the goal as a failing @next/playwright instant() e2e and work it to green, one verified route at a time; the shipped test then…

vercel/next.js · 170 tokens