Skill Test

Name: Skill Test
Author: gonimar

ASecurity

Validates Web Studio skills and agents: static (structural linter), spec (behavioural spec evaluation), category (rubric metrics), agent (agent spec evaluation), audit (coverage report). Uses the testing framework (catalog.yaml, quality-rubric.md, specs) from the kit repository or a project copy.

3 stars

0 votes

0 copies

0 views

Added 9/22/2026

ai-agentsgobashtesting

Security Analysis

A100/100

Scanned 9/22/2026

Install to Claude Code

$npx -y skills add gonimar/claude-web-studio --skill skill-test --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Skill Test?

Add the live security badge to your README — it updates automatically with every re-scan.

[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/gonimar-skill-test/badge)](https://www.skillsdirectory.com/skills/gonimar-skill-test)

More formats (shields.io, HTML) on the badges page.

Download with Pro

Files

SKILL.md

---
name: skill-test
description: "Validates Web Studio skills and agents: static (structural linter), spec (behavioural spec evaluation), category (rubric metrics), agent (agent spec evaluation), audit (coverage report). Uses the testing framework (catalog.yaml, quality-rubric.md, specs) from the kit repository or a project copy."
argument-hint: "static [name|all] | spec [name] | category [name|all] | agent [name|all] | audit"
user-invocable: true
allowed-tools: Read, Glob, Grep, Write, Bash, AskUserQuestion
model: sonnet
---

# Skill Test

Reply in the project conversation language (CLAUDE.md → Language); code, identifiers, paths and commit messages stay in English.

Tests the studio's own skills and agents (not the project). Framework directory — the first found of:
`./testing` (kit repository), `./web-studio-testing` (project copy), the plugin root's `testing/`
(from the session-start "Plugin root:" line or `claude plugin list --json`). Without a framework only `static` works.
Skill sources: `./skills/*/SKILL.md` (kit repo), `.claude/skills/*/SKILL.md` (copy mode) or the plugin root's `skills/`; agents likewise.

| Mode | What | Cost |
|---|---|---|
| `static [name\|all]` | 9 structural checks of SKILL.md | low |
| `spec [name]` | evaluate a skill against its behavioural spec | medium |
| `category [name\|all]` | category rubric metrics | low |
| `agent [name\|all]` | agent static checks + agent spec evaluation | medium |
| `audit` | coverage: who has a spec, last tested, result | low |

## Phase 1: Arguments
Parse mode and target; unknown → usage and stop. Read the framework's `catalog.yaml` (categories, spec paths, dates).

## Phase 2A: static — 9 SKILL.md checks
1. Frontmatter starts on line 1 with `---`; fields `name`, `description`, `argument-hint`, `user-invocable`, `allowed-tools` — FAIL if missing.
2. ≥ 2 phases (`## Phase N` or ≥ 2 `##`) — FAIL.
3. A verdict word (`PASS|FAIL|CONCERNS|APPROVED|ACCEPTED|PROPOSED|NEEDS REVISION|NEEDS CHANGES|BLOCKED|COMPLETE|READY|DONE|UPDATED|CLEAN|RELEASED|DEPLOYED|HARDENED|PLAYABLE|COMPLIANT|INITIALISED|RESOLVED|MITIGATED|WITHIN BUDGET|OVER BUDGET|ON TRACK|AT RISK|OFF TRACK|FIXED|IMPROVED`) — FAIL.
4. Ask-before-write: `May I write` (or an explicit gate sentence) when `Write|Edit` is in `allowed-tools` — FAIL; otherwise WARN. A gate that is not an `AskUserQuestion` with alternatives (coordination-rules, rule 7) — WARN. After the "write" answer: `touch .claude/.write-consent` (rule 7).
5. A "Next step" at the end — WARN; one that is not offered as an `AskUserQuestion` with alternatives — WARN.
6. A reference/template/rules link (`stack-reference/`, `templates/`, `rules/`) for authoring/analysis skills — WARN.
7. `argument-hint` non-empty and consistent with the argument-parsing phase — WARN.
8. Language: body in English, no project-specific or personal references (hostnames, names, private repo names) — WARN.
9. A "Reply in the project conversation language" line (the skill honours CLAUDE.md → Language regardless of its own English text) — WARN.
Output: a table of checks, `COMPLIANT | WARNINGS | NON-COMPLIANT`; for `all` — a summary table.

## Phase 2B: spec — behavioural evaluation
Read SKILL.md and the spec `skills/<category>/<name>.md`; for every case and assertion find the instructions in the skill text that satisfy it: PASS/FAIL/PARTIAL with a quoted line. Totals per case and protocol. "May I write the result to `results/<name>-<date>.md` and update `catalog.yaml`?" as one `AskUserQuestion`: results and catalog (Recommended) · results only · do not write.

## Phase 2C: category — rubric
The category section of `quality-rubric.md` → each metric PASS/WARN/FAIL with justification.

## Phase 2D: agent
Static: the agent file exists, `name/description/model/tools`, the collaboration protocol block, a stack-reference link, domain and "never"/escalation described. Then evaluate against `agents/<tier>/<name>.md` (5 cases) as in 2B.

## Phase 2E: audit
A table of all skills/agents: category/tier, spec present?, last_static/spec/category (date, result), priority. Uncatalogued files listed separately.

Verdict: `COMPLIANT` | `WARNINGS` | `NON-COMPLIANT`. Next step — one `AskUserQuestion`: `/skill-improve <name>` for failures (Recommended) · `/skill-test spec <name>` next · stop here.

Attribution

gonimar

View source More from gonimar →

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

--- name: skill-test description: "Validates Web Studio skills and agents: static (structural linter), spec (behavioural spec evaluation), category (rubric metrics), agent (agent spec evaluation), audit (coverage report). Uses the testing framework (catalog.yaml, quality-rubric.md, specs) from the kit repository or a project copy." argument-hint: "static [name|all] | spec [name] | category [name|all] | agent [name|all] | audit" user-invocable: true allowed-tools: Read, Glob, Grep, Write, Bash, AskUserQuestion model: sonnet --- # Skill Test Reply in the project conversation language (CLAUDE.md → Language); code, identifiers, paths and commit messages stay in English. Tests the studio's own skills and agents (not the project). Framework directory — the first found of: `./testing` (kit repository), `./web-studio-testing` (project copy), the plugin root's `testing/` (from the session-start "Plugin root:" line or `claude plugin list --json`). Without a framework only `static` works. Skill sources: `./skills/*/SKILL.md` (kit repo), `.claude/skills/*/SKILL.md` (copy mode) or the plugin root's `skills/`; agents likewise. | Mode | What | Cost | |---|---|---| | `static [name\|all]` | 9 structural checks of SKILL.md | low | | `spec [name]` | evaluate a skill against its behavioural spec | medium | | `category [name\|all]` | category rubric metrics | low | | `agent [name\|all]` | agent static checks + agent spec evaluation | medium | | `audit` | coverage: who has a spec, last tested, result | low | ## Phase 1: Arguments Parse mode and target; unknown → usage and stop. Read the framework's `catalog.yaml` (categories, spec paths, dates). ## Phase 2A: static — 9 SKILL.md checks 1. Frontmatter starts on line 1 with `---`; fields `name`, `description`, `argument-hint`, `user-invocable`, `allowed-tools` — FAIL if missing. 2. ≥ 2 phases (`## Phase N` or ≥ 2 `##`) — FAIL. 3. A verdict word (`PASS|FAIL|CONCERNS|APPROVED|ACCEPTED|PROPOSED|NEEDS REVISION|NEEDS CHANGES|BLOCKED|COMPLETE|READY|DONE|UPDATED|CLEAN|RELEASED|DEPLOYED|HARDENED|PLAYABLE|COMPLIANT|INITIALISED|RESOLVED|MITIGATED|WITHIN BUDGET|OVER BUDGET|ON TRACK|AT RISK|OFF TRACK|FIXED|IMPROVED`) — FAIL. 4. Ask-before-write: `May I write` (or an explicit gate sentence) when `Write|Edit` is in `allowed-tools` — FAIL; otherwise WARN. A gate that is not an `AskUserQuestion` with alternatives (coordination-rules, rule 7) — WARN. After the "write" answer: `touch .claude/.write-consent` (rule 7). 5. A "Next step" at the end — WARN; one that is not offered as an `AskUserQuestion` with alternatives — WARN. 6. A reference/template/rules link (`stack-reference/`, `templates/`, `rules/`) for authoring/analysis skills — WARN. 7. `argument-hint` non-empty and consistent with the argument-parsing phase — WARN. 8. Language: body in English, no project-specific or personal references (hostnames, names, private repo names) — WARN. 9. A "Reply in the project conversation language" line (the skill honours CLAUDE.md → Language regardless of its own English text) — WARN. Output: a table of checks, `COMPLIANT | WARNINGS | NON-COMPLIANT`; for `all` — a summary table. ## Phase 2B: spec — behavioural evaluation Read SKILL.md and the spec `skills/<category>/<name>.md`; for every case and assertion find the instructions in the skill text that satisfy it: PASS/FAIL/PARTIAL with a quoted line. Totals per case and protocol. "May I write the result to `results/<name>-<date>.md` and update `catalog.yaml`?" as one `AskUserQuestion`: results and catalog (Recommended) · results only · do not write. ## Phase 2C: category — rubric The category section of `quality-rubric.md` → each metric PASS/WARN/FAIL with justification. ## Phase 2D: agent Static: the agent file exists, `name/description/model/tools`, the collaboration protocol block, a stack-reference link, domain and "never"/escalation described. Then evaluate against `agents/<tier>/<name>.md` (5 cases) as in 2B. ## Phase 2E: audit A table of all skills/agents: category/tier, spec present?, last_static/spec/category (date, result), priority. Uncatalogued files listed separately. Verdict: `COMPLIANT` | `WARNINGS` | `NON-COMPLIANT`. Next step — one `AskUserQuestion`: `/skill-improve <name>` for failures (Recommended) · `/skill-test spec <name>` next · stop here.