terminal-bench-loop

Run one Terminal-Bench task through a bounded Paperclip smoke/diagnosis/fix loop. Use when asked to drive Terminal-Bench until it passes, rerun a Terminal-Bench loop, or iterate with board-gated fixes and diagnosis.

Quality

88%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

77%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is highly actionable with a clear, checkpointed multi-step workflow, but it is verbose with restated content and relies on an external doc/execution-semantics.md reference that is not bundled with the skill.

Suggestions

Either bundle the referenced doc/execution-semantics.md (and move the relevant state-machine detail there) or inline the essential execution-semantics terms and remove the dangling reference, so progressive disclosure is one level deep with real files.

Consolidate the restated content — the worktree rule, liveness rule, and pitfalls repeat procedure points; fold the unique parts into the relevant steps and trim the rest to improve conciseness.

Reduce repetition of the three invariants across the invariants section, the procedure, and the pitfalls to tighten the token budget.

Dimension	Reasoning	Score
Conciseness	The ~220-line body is dense and largely necessary, but the invariants, liveness rule, worktree rule, and pitfalls restate points already made in the procedure, so it could be tightened. It is mostly efficient but includes repetition that could be trimmed.	2 / 3
Actionability	Concrete, executable guidance throughout — exact env vars (PAPERCLIPAI_CMD, PAPERCLIP_HARBOR_RUNNER_CONFIG), exact idempotency keys, exact issue titles, status transitions, and a runnable smoke command (pnpm smoke:terminal-bench-loop-skill). Absence of code is fine for this instruction-only operational skill.	3 / 3
Workflow Clarity	Steps 0–9 are clearly sequenced with explicit validation checkpoints (verification checklist, stop rules) and feedback loops (diagnose → fix → rerun → re-diagnose) for a risky control-plane operation.	3 / 3
Progressive Disclosure	The body repeatedly points to doc/execution-semantics.md as a separate document to read, but no references/, scripts/, assets/, or doc/ directory exists in the bundle, so the referenced external material is missing and the main file is a large monolith.	2 / 3
	Total	10 / 12 Passed

Description

100%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description is concise, third-person, and covers both what the skill does and when to use it with natural trigger phrasing. It is specific and clearly distinguishable from other skills.

Dimension	Reasoning	Score
Specificity	Names multiple concrete actions — 'Run one Terminal-Bench task through a bounded Paperclip smoke/diagnosis/fix loop' specifies running, smoke, diagnosis, and fix phases rather than vague language.	3 / 3
Completeness	Explicitly answers both what ('Run one Terminal-Bench task through a bounded... loop') and when ('Use when asked to drive Terminal-Bench until it passes, rerun...'), with explicit triggers.	3 / 3
Trigger Term Quality	Natural request phrasings users would say are present — 'drive Terminal-Bench until it passes', 'rerun a Terminal-Bench loop', 'iterate with board-gated fixes and diagnosis' — covering common variations of the request.	3 / 3
Distinctiveness Conflict Risk	A narrow niche (Terminal-Bench loop via Paperclip with board-gated fixes) with distinct triggers makes it unlikely to fire for an unrelated skill.	3 / 3
	Total	12 / 12 Passed

Validation

100%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: paperclipai/paperclip
Path: .agents/skills/terminal-bench-loop/SKILL.md
Commit: 14f20be

Reviewed: 4 days ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.