PP Agent Toolkit

Fix Hack Learn Week · Microsoft CAPE + CAT

Inspect.
Validate.
Ship.

One workflow for agent quality, risk visibility, and delivery confidence — built for Copilot Studio.

76 tests passing
v0.4.0 current release
MIT open source
7
Capability pillars
5
Check categories
76
Automated tests
MIT
Open source license

Problem → Solution

From scattered signals
to decision-ready clarity

⚠ Before

Signals scattered across exports and tools
Dependency and conversation insights hard to surface
Instruction quality checks inconsistent
Go/no-go decisions arrive late

✓ After

Unified ingestion into one analysis workflow
Visual maps reveal bottlenecks instantly
Model-aware validation with transparent results
Decision-ready confidence with safe export flow

Animated Flow

Six steps from upload
to confidence

01

Ingest

Load solution ZIP, snapshot, or transcript

02

Detect

Classify artifacts and extract components

03

Visualize

Generate topic and dependency maps

04

Validate

Run instruction and security quality gates

05

Analyse

Score behavior and test-coverage fitness

06

Export

Ship a safe renamed copy with confidence

Capabilities

Seven actions.
One unified toolkit.

🗺

Visualize

Graph-first structure maps of any solution

Solution ZIP

Validate

Model-aware instruction quality checks

Snapshot ZIP
🔍

Check

20+ security and compliance rules

Solution ZIP
📊

Analyse

Conversation analytics with latency insights

Transcript JSON
🧪

Evals

Score and improve test coverage fitness

Solution ZIP
🔗

Dependencies

Visual dependency resolution at solution level

Any ZIP

Rename

Safe duplicate with publisher prefix rewriting

Solution ZIP

All in one

Single entry point for all workflows and formats

v0.4.0

Workflow

Four moves to ship.

01

Upload

Solution or snapshot ZIP, or transcript JSON

02

Inspect

Instant structural and risk visibility

03

Harden

Score, improve, and preview before export

04

Export

Ship an import-safe copy in one step

Architecture

Lean stack,
high signal.

Technical stack

  • Web UI — Reflex single-page interface
  • CLI — Typer-based rename and fetch flows
  • Models — Pydantic v2 data contracts
  • Parsing — hardened ZIP, XML, and YAML
  • Deployment — containerized runtime

Business impact

  • Reduce manual review with structured diagnostics
  • Increase delivery consistency across projects
  • Shift quality checks earlier in the lifecycle
  • Lower risk of overwrite and configuration drift
  • Improve confidence before environment import

Quality Signals

Confidence you can demo.

🚀

Release maturity

v0.4.0 with dedicated release notes and summary

🛡

Quality gates

Linting, formatting, and 76 automated tests

🔒

Security posture

Prompt-injection and hardcoded-credential checks

Open source trust

MIT licensed with issue and feature request channels

Dev Instance

Need access to the public dev tools instance?

Next Step

Agent quality as a
repeatable system.

Use this as your walkthrough layer for demos, internal alignment, and go/no-go delivery reviews.