n-agentic-harnesses-anthropic

Design, evaluate, and improve agentic harnesses — the orchestration layer around LLM-powered tools, agents, assistants, copilots, workflow runtimes, and AI-driven product features. Use this skill whenever the user mentions building an agentic system, structuring tool use, adding permissions or approval gates, designing multi-step AI workflows, managing context windows or memory, making agents durable or resumable, evaluating or pressure-testing an existing harness, planning phased implementation for an AI product, reviewing agent architecture, improving agent UX or observability, or asking how to know if their harness is actually good. Also use when the user describes problems that imply harness gaps — like agents doing unexpected things, context getting stale, sessions not surviving crashes, tools running without permission, or costs spiraling — even if they do not use the word "harness."

Quality

85%

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

Quality

Content

75%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 well-structured, router-style skill body that assumes Claude's intelligence, classifies requests into clear modes, and routes to a one-level-deep reference set with strong navigation. Main weaknesses are minor reference-list redundancy and the inability to verify the referenced bundle files exist.

Suggestions

De-duplicate the reference listings: Step 1 "Default reads" could reference the Step 3 catalog rather than re-listing paths, tightening conciseness.

Add one short worked-example output (a sample findings list or phased plan) inline so Claude has a concrete template before deferring to references/10.

Confirm the 11 referenced files ship alongside SKILL.md so the index's navigation links resolve for users.

Dimension	Reasoning	Score
Conciseness	The body is lean and router-style, assuming Claude's competence (no explanation of what agents/LLMs/harnesses are), but reference paths appear redundantly in both Step 1 "Default reads" and the full Step 3 listing, a trimming opportunity that keeps it just below anchor 5.	4 / 5
Actionability	Concrete, mode-specific routing (exact reference files per mode), a product-shape taxonomy, and explicit output-contract deliverables give actionable guidance; the gap is the absence of a worked example output, with examples deferred to references/10.	4 / 5
Workflow Clarity	Clear sequence (Step 0 gather context -> Step 1 classify request -> Step 2 classify shape -> Step 3 read references -> Output Contract -> Final Check) with a Step 0 clarification gate and a "Final Check Before Responding" checklist; not a clean 5 because there are no mid-process error-recovery feedback loops, though none are strictly required for this non-destructive routing skill.	4 / 5
Progressive Disclosure	The body is explicitly an index ("This file is the index", "Read only the files you need", "Do not rely on reference-to-reference chains") with 11 one-level-deep references each carrying a clear "Read when..." signal — excellent navigation design; scored 4 rather than 5 because the referenced files were not present in the review bundle to verify, leaving a minor completeness gap.	4 / 5
	Total	16 / 20 Passed

Description

95%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, well-constructed description that explicitly states capabilities, defines the domain precisely, and provides extensive natural trigger language including problem-statement phrasings for users who won't say "harness." Minor specificity gap is the high-level action verbs.

Suggestions

Consider sharpening the opening verbs beyond "design, evaluate, improve" with one or two more concrete operations (e.g., "route, sequence, and pressure-test") to lift specificity toward anchor 5.

The description is long; a minor trim of overlapping trigger clauses could improve token efficiency without losing coverage.

Dimension	Reasoning	Score
Specificity	Names the domain and three concrete action verbs ("Design, evaluate, and improve agentic harnesses") plus an extensive elaboration of in-scope subsystems, but the core verbs are somewhat high-level compared to anchor-5's granular operations, leaving minor coverage gaps.	4 / 5
Completeness	Explicitly answers both what ("Design, evaluate, and improve agentic harnesses") and when ("Use this skill whenever..." and "Also use when...") with concrete trigger phrases.	5 / 5
Trigger Term Quality	Comprehensive natural-language triggers including "building an agentic system", "structuring tool use", "adding permissions or approval gates", "managing context windows or memory", plus problem-statement phrasings ("agents doing unexpected things", "costs spiraling") and an explicit "even if they do not use the word 'harness'" clause covering synonyms.	5 / 5
Distinctiveness Conflict Risk	Carves a clear niche ("agentic harnesses — the orchestration layer around LLM-powered tools, agents, assistants, copilots, workflow runtimes") with distinct triggers and minimal conflict risk against unrelated skills.	5 / 5
	Total	19 / 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: 20 missing	Warning

	Total	15 / 16 Passed

Repository: NateBJones-Projects/OB1
Path: skills/n-agentic-harnesses/variants/anthropic/SKILL.md
Commit: e79e9cb

Reviewed: 11 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.