eval-harness

克劳德代码会话的正式评估框架，实施评估驱动开发（EDD）原则

Quality

44%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Medium

Suggest reviewing before use

Fix and improve this skill with Tessl

tessl review fix ./docs/zh-CN/skills/eval-harness/SKILL.md

Quality

Content

57%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.

The body is actionable and reasonably concise, with real commands and reusable templates, but repeats grader/eval-type content across sections and lacks explicit validation feedback loops in its workflow. As a single inline file with no progressive reference structure, it sits mid-rubric.

Suggestions

Add an explicit validate→fix→retry loop in the 评估 step (e.g., on FAIL, revise and re-run until pass@k threshold is met) to lift workflow clarity.

Consolidate the duplicated grader/eval-type material from '评分器类型' and '产品评估 (v1.8)' into one section, or move the bulk into a references file referenced one level deep.

Fix the template typo ('标准标准 3') and replace placeholder lines like '[Run each capability eval, record PASS/FAIL]' with a concrete runnable command.

Dimension	Reasoning	Score
Conciseness	The body is largely template-driven and assumes Claude's competence, but it repeats grader/eval-type material across '评估类型', '评分器类型', and '产品评估' sections, and templates carry minor padding and a typo ('标准标准 3').	3 / 5
Actionability	It provides concrete, executable bash commands (grep, npm test, npm run build), slash-command invocations, and copy-pasteable markdown templates that cover common cases, with only minor placeholder gaps like '[Run each capability eval, record PASS/FAIL]'.	4 / 5
Workflow Clarity	A clear four-step sequence (定义→实现→评估→报告) exists, but the eval/report steps lack explicit validate→fix→retry feedback loops; for a batch-evaluation skill the missing recovery checkpoints caps workflow clarity at 3.	3 / 5
Progressive Disclosure	The file is well-sectioned but all content lives inline in a single ~300-line SKILL.md with no bundle reference files; eval-type and grader-type material that could be split out remains inline, so structure is present but not optimally separated.	3 / 5
	Total	13 / 20 Passed

Description

32%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.

The description identifies a clear niche but is too abstract: it states the concept (EDD framework) without concrete capabilities and lacks any 'when to use' trigger guidance. It is distinguishable but weak on completeness and trigger terms.

Suggestions

Add concrete capabilities, e.g. '定义能力评估与回归评估，用代码/模型/人工评分器运行 pass@k 与 pass^k 指标，生成评估报告'.

Append an explicit trigger clause such as 'Use when 为 Claude Code 任务设定完成标准、创建回归测试套件、或跨模型版本基准测试代理性能'.

Include natural user-facing terms and synonyms (评估, 回归测试, pass@k, 基准测试) so the skill triggers on realistic phrasing.

Dimension	Reasoning	Score
Specificity	The description names the domain ('正式评估框架' for Claude Code sessions) and one abstract action ('实施评估驱动开发（EDD）原则') but lists no concrete, granular capabilities; actions are minimal and high-level rather than enumerated.	2 / 5
Completeness	It gives a vague 'what' (a formal eval framework implementing EDD) and no explicit 'when' guidance; with no 'Use when...' trigger the completeness is capped and sits below the midpoint.	2 / 5
Trigger Term Quality	It uses domain terms ('评估框架', '评估驱动开发') but lacks the natural phrases a user would say when needing the skill and includes no 'Use when...' trigger clause or synonyms/file extensions.	2 / 5
Distinctiveness Conflict Risk	The niche — a formal evaluation-driven-development framework for Claude Code sessions — is specific and unlikely to overlap with unrelated skills, with only minor overlap risk against generic testing skills.	4 / 5
	Total	10 / 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: affaan-m/ECC
Path: docs/zh-CN/skills/eval-harness/SKILL.md
Commit: e04ea0b

Reviewed: 2 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.