write-agent-benchmarks

Create, maintain, and run evidence-based benchmarks for AI agents. Use when setting up testing infrastructure, writing new test scenarios, or evaluating agent performance.

Quality

75%

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

Fix and improve this skill with Tessl

tessl review fix ./plugins/flowai-devtools/skills/write-agent-benchmarks/SKILL.md

Quality

Content

67%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-organized, largely actionable skill body with clear sequenced workflows and explicit validation gates in its debugging guidance. Its main weaknesses are padded conceptual sections that hurt token efficiency, and two broken/orphaned reference paths that weaken progressive disclosure.

Suggestions

Tighten the "Context & Philosophy" section: convert the principle prose and "Bad/Good" restatements into lean directives so every token earns its place (addresses conciseness).

Fix the broken §9 link to `acceptance-tests/config.json` (the file is absent from the bundle) and add a link to the orphaned `references/PROMPTS.md` from the Judge section so every bundle file is reachable (addresses progressive_disclosure).

Add an explicit validation checkpoint to the "Creating a New Benchmark" workflow — e.g. verify the scenario registers and passes a dry-run before Step 5 is considered complete (addresses workflow_clarity).

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient domain-specific material (mockbin/PATH-shadow, trigger-scenario contracts) Claude would not already know, but the "Context & Philosophy" section and some prose (e.g. the "Bad/Good" principle restatements, section 8 Configuration Principles) are padded conceptual explanation that could be tightened. It is above 2 because it does not explain basic concepts Claude knows (Docker, PDFs), but below 4 due to these unnecessary explanatory sections.	3 / 5
Actionability	It provides mostly executable guidance — a real TypeScript `mod.ts` template, checklist-contract code, the universal result JSON schema, and concrete debug commands (`ls <workDir>/mockbin/`, grep for the sentinel token). It is not a 5 because the template carries unfilled placeholders (`<skill-id>`, `<natural user query>`) and the scenario example is pseudocode, leaving minor gaps.	4 / 5
Workflow Clarity	Both workflows (Creating a New Benchmark; Running & Debugging) are clearly sequenced, and the Debugging section has explicit validation gates ("Only after steps 1–4 pass is it safe to edit SKILL.md") plus a RED-phase feedback loop. It is not a 5 because the "Creating a New Benchmark" workflow (Steps 1–5) lacks an explicit validation checkpoint, leaving a minor validation gap.	4 / 5
Progressive Disclosure	The document is well-structured with numbered sections and a clearly signaled §9 Assets & References using markdown links. It is not a 5 because §9 links a non-existent `acceptance-tests/config.json`, the bundle's `references/PROMPTS.md` is orphaned (never linked from the body), and the lengthy §6.1–6.3 detail is inlined rather than split into one-level-deep reference files.	4 / 5
	Total	15 / 20 Passed

Description

83%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, third-person description that concretely states what the skill does and gives an explicit "Use when" trigger clause with natural phrasing. Its only weakness is minor overlap risk between general testing vocabulary and its specific agent-benchmarking niche.

Dimension	Reasoning	Score
Specificity	"Create, maintain, and run evidence-based benchmarks" names the domain plus three concrete verbs, and the when-clause adds "setting up testing infrastructure, writing new test scenarios, or evaluating agent performance" — several specific actions with only minor granularity gaps. It is above anchor 3 (which allows just 1-2 actions) but the actions are higher-level lifecycle verbs rather than the granular operations in the anchor-5 example, so it stops at 4.	4 / 5
Completeness	It explicitly answers both: what ("Create, maintain, and run evidence-based benchmarks for AI agents") and when ("Use when setting up testing infrastructure, writing new test scenarios, or evaluating agent performance") with concrete trigger phrases, matching the anchor-5 example structure. It is above 4 because the when-clause is explicit and trigger-phrase-based rather than weakly implied.	5 / 5
Trigger Term Quality	Natural terms users would say are present — "benchmarks", "testing infrastructure", "test scenarios", "evaluating agent performance", "AI agents" — giving good keyword coverage. It is not a 5 because a few natural synonyms/variations (e.g. "agent evaluation", "CI/regression testing") are missing and there are no file-extension triggers, though those are less applicable to this non-file-format skill.	4 / 5
Distinctiveness Conflict Risk	The "AI agents" / "benchmarks" / "agent performance" framing carves a clear niche with distinct triggers. It is not a 5 because "testing infrastructure" and "test scenarios" could overlap with broader code-testing/dev skills, giving minor conflict risk with closely related skills.	4 / 5
	Total	17 / 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: 1 missing	Warning
referenced_paths_exist	Referenced path issues: 1 missing	Warning

	Total	14 / 16 Passed

Repository: korchasa/flowai-plugins
Path: plugins/flowai-devtools/skills/write-agent-benchmarks/SKILL.md
Commit: b4ba256

Reviewed: ~22 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.