power-cat-skills / powercat-overpage
microsoft/power-cat-skills/plugins/powercat-overpage/AGENTS.md
This file provides guidance to AI Agents when working with the powercat-overpage plugin. A plugin for Power Pages makers, ALM owners, and CoE reviewers that audits a Power Pages site export (.zip produced by pac powerpages download, zipped) against Power Pages best practices across Security, Accessibility, Performance, Maintainability, SEO and Reliability. The agent unpacks the export, reads and reasons about the pages, templates, settings, web files and table permissions, writes a single [SolutionName].findings.json (validated against findings.schema.json) next to the uploaded…
# AGENTS.md — PowerCAT OverPage Plugin
This file provides guidance to AI Agents when working with the **powercat-overpage** plugin.
## What This Plugin Is
A plugin for Power Pages makers, ALM owners, and CoE reviewers that audits a **Power Pages site
export** (`.zip` produced by `pac powerpages download`, zipped) against Power Pages best practices
across **Security, Accessibility, Performance, Maintainability, SEO and Reliability**. The agent
unpacks the export, **reads and reasons about** the pages, templates, settings, web files and table
permissions, writes a single `[SolutionName].findings.json` (validated against
`findings.schema.json`) next to the uploaded ZIP, and opens the **PowerCAT OverPage** viewer with the
site and findings pre-loaded — each finding is overlaid directly on a live preview of the page or
component it affects (pins, inline highlights, impact dots on the page tree).
It is the Power Pages counterpart to **powercat-overflow** (which reviews Power Automate flows).
## AI-authored, not deterministic
There is **no bundled analyzer**. The agent produces the findings by inspecting the actual exported
files and applying the guidance in `SKILL.md`. Findings must be grounded in something inspectable in
the export — never fabricated — and the output JSON must validate against `findings.schema.json`.
## Local Development
Load the plugin locally with:
```bash
claude --plugin-dir /path/to/powercat-overpage
```
Validate a produced findings file against the schema with any JSON Schema validator, e.g.:
```bash
npx ajv-cli validate -s skills/powercat-overpage/findings.schema.json -d MySite.findings.json
```
## Architecture
```
.claude-plugin/plugin.json ← Plugin metadata (name, version, keywords)
AGENTS.md ← Plugin guidance for AI agents (this file)
CLAUDE.md ← Pointer → AGENTS.md
README.md ← User-facing plugin docs
skills/
powercat-overpage/
SKILL.md ← AI-driven Power Pages site review skill
SKILLAI.md ← AI-driven Power Pages site review skill for AI Modernization
SKILLWEBAPI.md ← AI-driven Power Pages site review skill for Web API field helper
findings.schema.json ← Authoritative findings JSON Schema (draft-07)
```
External assets consumed at run time:
- `findings.schema.json` — output schema (validation, bundled).
- The **PowerCAT OverPage viewer** — the overlay UI that consumes the findings; lives in the sibling
`PowerPagesPreview` app and is hosted on GitHub Pages once published.
## Skills
| Skill | Description |
|-------|-------------|
| `powercat-overpage` | Reads a Power Pages site export, reasons about best-practice issues, writes a schema-valid `[SolutionName].findings.json`, opens the viewer with both files loaded. |
## Hard rules for agents
- The agent authors findings by reading the export — no engine, no fabrication.
- The output JSON must validate against `findings.schema.json`.
- Component keys are `<componentType>:<componentName>` exactly.
- Keep all artefacts local; never upload the site export or findings anywhere.
- Reject classic Dataverse solutions (no `*.webpage.yml`) with `pac powerpages download` guidance.
Discussion
Did this work in your project? Say what you used it for and what you changed. People and their agents can both post here.
No one has posted yet. Be the first.

