agent-harness

Turn any domain folder of skills into a bounded agentic loop: compile a goal into a verifiable task plan, execute tasks with the domain's own tools, verify every task with machine-run checks, retry with caps, escalate to a human when budgets exhaust, and refuse to close until everything is verified or explicitly waived. Use when you want an agent or subagent to pick up a goal and drive it to a verified close across one of this repo's 18 domains ('run this goal through the engineering harness', 'set up an agentic loop for marketing work', 'make the finance domain self-verifying'). NOT for authoring Claude Code Workflow-tool .js scripts (workflow-builder), N-agent tournaments on one task (agenthub), single-file metric optimization (autoresearch-agent), or discovering published loop recipes (loop-library).

Quality

96%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

—

The risk profile of this skill

Quality

Content

92%Weight 40%Scale 1-5

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

A tight, highly operational SKILL.md body: executable quick-start, explicit validation feedback loops, and well-signaled one-level references into a real bundle. The only weakness is mild verbosity in the stylistic opener and the related-skills survey.

Suggestions

Tighten the opening ('You are a harness operator, not a hero...') to one operational sentence; the contract ASCII diagram largely duplicates the three-layer paragraph that follows it.

Trim the Related skills list to the two or three most likely to be confused with this skill, or move the full mapping into references/domain_harness_design.md which already covers the reuse map.

Dimension	Reasoning	Score
Conciseness	Mostly lean and operational with no concept-explanation padding, but the stylistic opener ('You are a harness operator, not a hero'), the ASCII contract diagram, and the six-entry Related skills list are minor tokens that could be trimmed. Not a 5 because a few sections don't purely earn their place.	4 / 5
Actionability	Quick start gives fully executable copy-paste commands with exact arguments and flags, the exit-code table maps mechanically to branch decisions, and Hard rules cite specific exit codes (3, 4, 5, 6) — copy-paste ready across the common cases.	5 / 5
Workflow Clarity	A clearly sequenced 0→1→2→3→4 workflow with explicit validation checkpoints and feedback loops: verify runs checks via subprocess, a passing record without evidence is rejected, verify failure consumes an attempt, and close is refused (exit 4) while any task is unverified — exactly the validate→fix→retry pattern the rubric rewards.	5 / 5
Progressive Disclosure	The body is an overview pointing to three real one-level-deep reference files (verification_discipline.md, agentic_loop_canon.md, domain_harness_design.md) via clearly signaled markdown links, with scripts and assets appropriately separated — easy navigation, no nested reference chains.	5 / 5
	Total	19 / 20 Passed

Description

100%Weight 40%Scale 1-5

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

An exemplary description: concrete capabilities, natural multi-domain trigger phrases, explicit what/when guidance, and a clear negative-boundary list distinguishing it from sibling skills. Voice is third person and free of fluff.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions spanning the loop lifecycle — 'compile a goal into a verifiable task plan', 'execute tasks with the domain's own tools', 'verify every task with machine-run checks', 'retry with caps', 'escalate to a human', 'refuse to close until everything is verified or explicitly waived' — comprehensive coverage matching the anchor.	5 / 5
Completeness	Explicitly answers both 'what' (the full first sentence of capabilities) and 'when' ('Use when you want an agent or subagent to pick up a goal and drive it to a verified close across one of this repo's 18 domains'), with concrete trigger phrases.	5 / 5
Trigger Term Quality	Provides natural user utterances as triggers ('run this goal through the engineering harness', 'set up an agentic loop for marketing work', 'make the finance domain self-verifying') across multiple domains, with good synonym/variation coverage rather than only jargon.	5 / 5
Distinctiveness Conflict Risk	A clear niche with an explicit 'NOT for...' clause naming four sibling skills (workflow-builder, agenthub, autoresearch-agent, loop-library), minimizing conflict risk.	5 / 5
	Total	20 / 20 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
referenced_paths_exist	Referenced path issues: 1 deeper-than-1-level	Warning

	Total	15 / 16 Passed

Repository: alirezarezvani/claude-skills
Path: engineering/agent-harness/skills/agent-harness/SKILL.md
Commit: 19392f7

Reviewed: ~3 hours 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.