Burrow Watch

This site is shared with workshop participants and program partners. Enter the password to continue.

Workshop toolkit

This page is different from the resources page. Resources are Burrow Watch's own materials — the program's background, transcripts, and data. This page is about the thing built on top of them: a configured Claude Project that turns those materials into coded findings, briefs, and reports. Here's how it works, how to run it in one go, and how to build the same thing in Copilot.

Assistant, or agent?

Both words get used loosely, and it's worth being precise before you build either one.

An AI assistant, in the sense used throughout this project, is a chat set up with standing instructions and reference material, so it answers the same way every time rather than depending on how well someone happens to phrase a question. It waits for you to ask, answers, and stops. That's what a Claude Project or a Microsoft Copilot Agent Builder agent gives you: consistency and grounding, not independence.

An agent, more precisely, is something that can also act — call a tool, send a message, move a file — usually still with a person approving each step, and only occasionally running fully on its own schedule. The word gets applied to all three of these at once in most marketing, which is exactly why it's worth naming the difference out loud.

Configured Answers reliably using fixed instructions and knowledge. Doesn't act on anything. This is what the Burrow Watch Project's skills mostly do — coding transcripts, drafting a report.
Gated action Can call a tool or send something in the real world, but pauses for a person's approval first. The report-sender skill works this way — it drafts an email and stops before it actually sends.
Autonomous Runs on a schedule or trigger without anyone approving each step — a scheduled task, or a tool like n8n or Copilot Cowork. Powerful, and a deliberate choice to turn on, not a default.

Try it yourself

If you've set up a Claude Project with the instructions, knowledge files, and skills from below, these two prompts will run the pipeline end to end. Copy, paste, and Claude works through the steps on its own — checking email, coding transcripts, pulling in data, and drafting the report.

Quick start — one skill

Copied
Check my email for anything with "field notes" in the subject, pull in what you find, and code it together with our existing interview transcripts against our framework.

Full pipeline — everything at once

Copied
Using everything in this project: check my email for anything with "field notes" in the subject and pull in what's relevant, code that alongside our existing interview transcripts against our framework, give me the latest wombat conservation news brief, and then generate the full State of the Burrows report using all of that plus our quantitative data — including progress against each outcome in our theory of change. Don't send or share anything without asking me first.

These are deliberately simple, plain-language prompts — Claude works out which skill handles which part. The full live demo script (with narration, timing, and recovery lines) is a separate facilitator document, not published here.

Project instructions

Claude Project instructions

Paste this into a Project's instructions field — role, standing rules, and what the six skills are for.

Download Copied

Skills

Email intake

Searches connected email for field notes shared by rangers, the vet, or landholders.

Transcript coder

Codes interviews and field notes against the six-category framework.

Context researcher

Live web research on current wombat conservation news for the external brief.

Report generator

Assembles the quarterly report, including progress against theory-of-change outcomes.

Report sender

Drafts and sends the report by email, always with approval before it goes out.

Stakeholder brief

Builds a one-page briefing note ahead of a call with a named stakeholder.

Using this in Microsoft Copilot instead

Everything above is written for a Claude Project, but the underlying idea — instructions plus grounding material, applied consistently — carries over to Microsoft Copilot's Agent Builder. A few things to adjust along the way:

Instructions and knowledge — paste the Project instructions text into Agent Builder's instructions field the same way, and attach the background, transcripts, theory of change, and data files as knowledge sources.
No separate skill files — Agent Builder doesn't have an equivalent to individually installable Skills. Fold each skill's instructions into the main instructions field as separate labelled sections instead of six separate files.
Email intake — needs the Outlook/Microsoft 365 connector rather than Gmail. Read access works the same way; check current Copilot documentation for exact setup steps, since connector details change over time.
Sending the report — this is the one that doesn't carry over directly. Agent Builder can't take actions like sending mail on its own; that needs Copilot Studio's Actions layer instead. Plan to draft the email manually from Agent Builder's output, or build the sending step separately in Copilot Studio.
Running the "full pipeline" prompt — Agent Builder tends to work better with one instruction at a time rather than a long chained request. Try breaking the full-pipeline prompt above into its separate steps if it doesn't work as a single paste.