How do I install the How It Works skill?

Run npx skillmds@latest add tools-only/how-it-works-10 in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

What does the How It Works skill do?

Automated dashboard for comparing model performance across multiple GitLab CI runs. It is listed under Coding & Dev Tools on SkillMD.

Is How It Works safe to use?

This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: CAUTION, Skill Scanner: PASS. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with How It Works?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is How It Works free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published How It Works?

tools-only (@tools-only) published this skill. Their other Agent Skills are listed on their SkillMD profile.

How It Works by tools-only — Agent Skill

Name: How It Works
Author: tools-only

File contents

Overview

Automated dashboard for comparing model performance across multiple GitLab CI runs.

Features

Multi-model comparison
Trend analysis with Chart.js
Auto-discovery of recent evaluations
GitLab Pages integration
Judge mode support

How It Works

Run eval.py → Generate test results
Generate report → Create complete_eval_report.html
Extract metadata → save_report_metadata.py creates report_metadata.json
Aggregate → Fetch reports from multiple CI runs via GitLab API
Dashboard → Interactive comparison view with charts

Usage

Option A: Manual Mode (Specific Job IDs)

# In GitLab CI/CD Variables, set:
AGGREGATE_JOB_IDS=12345,12346,12347

# Trigger pipeline → Dashboard deployed to GitLab Pages

Option B: Auto Mode (Last N Jobs)

# Optional: Set limit (default: 10)
AGGREGATE_LIMIT=20

# Run scheduled or manual pipeline

Required Output Files

Your evaluation must produce:

complete_eval_report.html - Full HTML evaluation report
report_metadata.json - Extracted statistics (auto-generated)

Scripts Reference

Script	Purpose
`aggregate_reports.py`	Fetches reports from GitLab API, supports auto-discovery
`generate_dashboard.py`	Creates interactive HTML dashboard with Chart.js
`save_report_metadata.py`	Extracts metadata from HTML reports

Dashboard Features

Overview: Summary stats, best/worst runs
Models: Per-model performance comparison
Trends: Pass rate over time (separate line per model)
Components: Component-level accuracy breakdown
Failures: Detailed failure analysis with collapsible sections
Warnings: Tests with warnings (Judge mode)
Pipeline History: Clickable links to individual reports

Configuration Variables

Set in GitLab: Settings → CI/CD → Variables

Variable	Required	Default	Description
`AGGREGATE_JOB_IDS`	No	-	Comma-separated job IDs for manual mode
`AGGREGATE_LIMIT`	No	10	Number of recent jobs for auto mode
`AGGREGATION_TOKEN`	No	CI_JOB_TOKEN	Project access token for cross-pipeline access

View Dashboard

After aggregation completes:

https://<your-gitlab-instance>/<your-project>/pages/

Troubleshooting

No jobs found in auto mode

Verify generate_report job creates artifacts
Ensure at least one successful evaluation run exists

401 Unauthorized error

For cross-project aggregation, create Project Access Token
Set as AGGREGATION_TOKEN CI/CD variable with read_api scope

---name: how-it-works-10description: Automated dashboard for comparing model performance across multiple GitLab CI runs.---## OverviewAutomated dashboard for comparing model performance across multiple GitLab CI runs.### Features- Multi-model comparison- Trend analysis with Chart.js- Auto-discovery of recent evaluations- GitLab Pages integration- Judge mode support## How It Works1. **Run eval.py** → Generate test results2. **Generate report** → Create `complete_eval_report.html`3. **Extract metadata** → `save_report_metadata.py` creates `report_metadata.json`4. **Aggregate** → Fetch reports from multiple CI runs via GitLab API5. **Dashboard** → Interactive comparison view with charts## Usage### Option A: Manual Mode (Specific Job IDs)```bash# In GitLab CI/CD Variables, set:AGGREGATE_JOB_IDS=12345,12346,12347# Trigger pipeline → Dashboard deployed to GitLab Pages```### Option B: Auto Mode (Last N Jobs)```bash# Optional: Set limit (default: 10)AGGREGATE_LIMIT=20# Run scheduled or manual pipeline```## Required Output FilesYour evaluation must produce:- **`complete_eval_report.html`** - Full HTML evaluation report- **`report_metadata.json`** - Extracted statistics (auto-generated)## Scripts Reference| Script | Purpose || ------------------------- | -------------------------------------------------------- || `aggregate_reports.py` | Fetches reports from GitLab API, supports auto-discovery || `generate_dashboard.py` | Creates interactive HTML dashboard with Chart.js || `save_report_metadata.py` | Extracts metadata from HTML reports |## Dashboard Features- **Overview**: Summary stats, best/worst runs- **Models**: Per-model performance comparison- **Trends**: Pass rate over time (separate line per model)- **Components**: Component-level accuracy breakdown- **Failures**: Detailed failure analysis with collapsible sections- **Warnings**: Tests with warnings (Judge mode)- **Pipeline History**: Clickable links to individual reports## Configuration VariablesSet in GitLab: **Settings → CI/CD → Variables**| Variable | Required | Default | Description || ------------------- | -------- | ------------ | ---------------------------------------------- || `AGGREGATE_JOB_IDS` | No | - | Comma-separated job IDs for manual mode || `AGGREGATE_LIMIT` | No | 10 | Number of recent jobs for auto mode || `AGGREGATION_TOKEN` | No | CI_JOB_TOKEN | Project access token for cross-pipeline access |## View DashboardAfter aggregation completes:```https://<your-gitlab-instance>/<your-project>/pages/```## Troubleshooting### No jobs found in auto mode- Verify `generate_report` job creates artifacts- Ensure at least one successful evaluation run exists### 401 Unauthorized error- For cross-project aggregation, create **Project Access Token**- Set as `AGGREGATION_TOKEN` CI/CD variable with `read_api` scope

Frequently asked questions

How It Works

File contents

Overview

Features

How It Works

Usage

Option A: Manual Mode (Specific Job IDs)

Option B: Auto Mode (Last N Jobs)

Required Output Files

Scripts Reference

Dashboard Features

Configuration Variables

View Dashboard

Troubleshooting

No jobs found in auto mode

401 Unauthorized error

Related skills

Frequently asked questions

How do I install the How It Works skill?

What does the How It Works skill do?

Is How It Works safe to use?

Which AI agents work with How It Works?

Is How It Works free to use?

Who published How It Works?