Name: agent-harness
Rating: 71.2 (1 reviews)
Author: alirezarezvani

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

88%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

The canonical home for this skill is agent-harness in alirezarezvani/claude-skills

Quality

Content

88%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.

An exemplary operational skill body: lean, copy-paste-ready commands, a clearly sequenced loop with explicit validation and feedback, and well-signaled one-level-deep references. The main gap is that the referenced bundle files (references/, scripts/, assets/) are absent, breaking the progressive-disclosure navigation.

Suggestions

Ship the referenced bundle so the inline links resolve: create references/verification_discipline.md, references/agentic_loop_canon.md, references/domain_harness_design.md, the scripts/ entry points, and assets/harnesses/ manifests.

Verify the Quick start commands match the actual script CLIs (flag names, subcommands like init/next/record/verify/close) once scripts/ exists, so the copy-paste examples are truly executable.

Consider moving the 'Related skills' and 'Forcing questions' tables into a reference file to tighten the SKILL.md body further, since they are context rather than core procedure.

Dimension	Reasoning	Score
Conciseness	The body is information-dense with no padding or explanations of concepts Claude already knows (no 'what is a loop' filler); every section (contract, quick start, hard rules, tables) earns its place, though the forcing-questions and related-skills tables could be trimmed slightly.	4 / 5
Actionability	Quick start provides copy-paste-ready commands with full flags and expected JSON outputs shown inline as comments ('python3 scripts/goal_compiler.py --goal ... --manifest ... --out plan.json'), and the exit-code table gives mechanical branching guidance covering common cases.	5 / 5
Workflow Clarity	The multi-step process is explicitly sequenced (0→1→2→3→4) with validation checkpoints (verify runs checks via subprocess, close refused on unverified tasks, retry consumes an attempt with a changed approach, escalate on exhausted budgets) and a feedback loop shown in the --sample demo.	5 / 5
Progressive Disclosure	Structure is good with clear sections and one-level-deep, clearly signaled inline references ([references/verification_discipline.md], [references/agentic_loop_canon.md], [references/domain_harness_design.md]); docked from 5 because the referenced bundle directories (references/, scripts/, assets/) are not present in the reviewed bundle, so those navigational links are broken.	4 / 5
	Total	18 / 20 Passed

Description

87%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.

A strong, highly specific description with explicit 'what' and 'when' guidance, concrete trigger phrases, and clear negative-scoping against sibling skills. The only defect is second-person voice ('Use when you want...'), which the rubric penalizes on specificity.

Dimension	Reasoning	Score
Specificity	The description lists multiple concrete actions ('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'), which is comprehensive; docked one point because the 'Use when you want...' clause is second-person voice, which the rubric penalizes by reducing specificity by 1.	4 / 5
Completeness	It explicitly answers both 'what' (compile → plan → execute → verify → retry → escalate → close) 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, plus negative-scoping guidance.	5 / 5
Trigger Term Quality	It embeds natural user phrases in quotes ('run this goal through the engineering harness', 'set up an agentic loop for marketing work', 'make the finance domain self-verifying') giving good keyword coverage, but the primary vocabulary is technical ('bounded agentic loop', 'verifiable task plan') and a few common synonyms are missing.	4 / 5
Distinctiveness Conflict Risk	It carves a clear niche (bounded agentic loops across domains) and explicitly distinguishes from sibling skills (workflow-builder, agenthub, autoresearch-agent, loop-library), minimizing conflict risk.	5 / 5
	Total	18 / 20 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
relative_links	Relative link issues: 4 missing	Warning
referenced_paths_exist	Referenced path issues: 22 missing, 1 deeper-than-1-level	Warning

	Total	14 / 16 Passed

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

Reviewed: ~2 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.