CtrlK
BlogDocsLog inGet started
Tessl Logo

eval-harness

Formal evaluation framework for Claude Code sessions implementing eval-driven development (EDD) principles. Use when a Claude Code workflow needs a formal eval before it is trusted or changed.

60

Quality

70%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Medium

Suggest reviewing before use

Fix and improve this skill with Tessl

tessl review fix ./.agents/skills/eval-harness/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

61%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 well-structured with executable grader examples, eval templates, and a clear phased workflow, plus a worked authentication example. It would benefit from trimming pedagogical padding and adding an explicit eval failure feedback loop.

Suggestions

Add an explicit validate->fix->re-run feedback loop in the Evaluate workflow step (e.g., 'If any eval FAILS: fix the code, re-run the failing eval, repeat until all pass') to raise workflow clarity above 3.

Trim pedagogical padding in Philosophy, Metrics definitions, and Best Practices to assume Claude's competence and improve token efficiency.

Clarify or implement the '/eval define|check|report' integration commands, or mark them as conventional invocations rather than presenting them as ready-to-use without backing tooling.

DimensionReasoningScore

Conciseness

Mostly efficient with concrete templates and commands, but includes pedagogical padding Claude already knows (Philosophy's 'unit tests of AI development' framing, Metrics explaining 'At least one success in k attempts', aphoristic Best Practices like 'Slow evals don't get run').

3 / 5

Actionability

Provides concrete, executable guidance including shell graders ('grep -q ... && echo PASS', 'npm test -- --testPathPattern'), eval templates, and integration commands, with minor gaps from placeholder markdown templates and the invented '/eval' command that has no implementation.

4 / 5

Workflow Clarity

The Eval Workflow has a clear Define/Implement/Evaluate/Report sequence, but lacks an explicit validate->fix->re-run feedback loop for what is a repeated/batch eval operation; the missing feedback loop caps this at 3 per the rubric.

3 / 5

Progressive Disclosure

Single self-contained SKILL.md with no bundle files and well-organized section headers (When to Activate, Eval Types, Grader Types, Metrics, Eval Workflow, Example); good structure with minor gaps where the full template/example bulk could be trimmed.

4 / 5

Total

14

/

20

Passed

Description

78%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 clearly states both what the skill does and when to use it with a concrete 'Use when...' trigger, and uses third-person voice. It is somewhat abstract in naming capabilities and could broaden its natural trigger-term coverage with synonyms like 'benchmark' or 'regression test'.

Suggestions

Enumerate 2-3 concrete actions (e.g., 'define pass/fail criteria, measure pass@k reliability, generate regression reports') instead of the abstract phrase 'evaluation framework'.

Add natural synonyms to the trigger clause such as 'benchmarking agent reliability' or 'regression testing prompt or agent changes' to improve trigger-term coverage.

DimensionReasoningScore

Specificity

Names the domain ('Formal evaluation framework for Claude Code sessions implementing eval-driven development') and a concrete purpose ('needs a formal eval before it is trusted or changed'), but describes actions abstractly rather than enumerating specific concrete actions like define/measure/report.

3 / 5

Completeness

Explicitly answers both what ('Formal evaluation framework... implementing EDD principles') and when ('Use when a Claude Code workflow needs a formal eval before it is trusted or changed') with concrete trigger phrases.

5 / 5

Trigger Term Quality

Includes natural trigger phrasing ('Use when a Claude Code workflow needs a formal eval', 'before it is trusted or changed', 'eval'), but misses common synonyms users would naturally say such as 'benchmark', 'regression test', or 'pass@k'.

4 / 5

Distinctiveness Conflict Risk

Has a fairly clear niche (formal evals / EDD for Claude Code) but 'Claude Code workflow' is broad and could overlap with general testing or CI skills, so it is mostly distinct with minor overlap risk.

4 / 5

Total

16

/

20

Passed

Validation

93%

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

Validation15 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

allowed_tools_field

'allowed-tools' contains unusual tool name(s)

Warning

Total

15

/

16

Passed

Repository
affaan-m/ECC
Reviewed

Table of Contents

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.