Pi Extensible Workflows: Full Guide

You are helping me turn one specific YouTube video into real, durable learning.

Source video:
- Title: Pi Extensible Workflows: Full Guide
- URL: https://www.youtube.com/watch?v=qAiivspEHmU
- Topic: AI Strategy
- My current learning frame: Build a small 'develop until approved' workflow where a developer role implements a task in its own git worktree, a reviewer role critiques it with a capped number of retries, and a deterministic (non-agent) step merges and cleans up the work trees at the end.
- Why this matters: New playlist item from Andrea Baccega; queued for transcript-backed review, topic mapping, and a practical learning artifact.

Transcript anchors from this exact video:
- 0:54 / Evidence 1: "timestamp chapters if you want to skip along. Before going inside the the extension itself, I want you to introduce you to the workflows uh kind of concept which is uh which firstly introduced by the code team..."
- 4:16 / Evidence 2: "agents the uh end result do not end up on the main agent context window which is very very good compared to the sub aents that we were talking about before. You get a very very uh narrow..."
- 9:47 / Evidence 3: "this is what ends up to be inside the system prompt of the main agent I've been launching this from. So all of the work that has been done here is uh even if it was from an..."
- 13:43 / Evidence 4: "of concurrency that you want to for example parallelize um let's say that you have a very very high work parallel workflow uh where it spins up like um 15 different agents you want to limit up for..."
- 17:22 / Evidence 5: "whatever the the the workflow um requires some kind of external approval. Of course, you can use shell to for example wait for some kind of human interaction with some uh scripting or whatever or another binary whatever..."
- 21:29 / Evidence 6: "the summary sorry the the summarizer model the summarizer agent because it gets inside the system prompt. Uh I I'm still experimenting of what to put here but yeah you get the idea. You we can see another..."
- 23:32 / Evidence 7: "only on the task because you removed all the noise from the system prompt and the and the context itself. So um yeah here it is talking how you can do stuff for example you can even modify..."

Video-aware target:
- Prompt lane: Agent harness
- Mechanism to extract: Identify what surrounding harness makes the model more useful than chat alone.
- Artifact to produce: A one-page agent harness map with tool boundaries, state ownership, and proof signals.
- Artifact must include: model role; tools; state/memory; permission boundary; verification proof

Your task:
1. Use the transcript anchors above as the primary source packet. If you add outside context, label it clearly as outside context and keep it secondary.
2. Create a source-check table with columns: timestamp, claim, transcript support, what the demo proves, confidence, and what still needs verification.
3. Extract the actual teachable mechanism from the video: Identify what surrounding harness makes the model more useful than chat alone. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A one-page agent harness map with tool boundaries, state ownership, and proof signals.
5. Include:
   - a plain-English definition of the core idea
   - a diagram or structured model using this sequence: User intent -> Model role -> Tool surface -> State and memory -> Verification loop -> Reusable operating rule
   - answers to these source questions: What does the video claim the agent can do? | What surrounding system makes that claim plausible? | What proof is shown instead of merely asserted?
   - 3 concrete examples that apply the video idea to real agentic work, such as a repo-editing harness; a local research assistant; a recurring refresh agent
   - 2 failure modes the video helps prevent, chosen from the transcript evidence and these likely risks: treating model choice as architecture; ignoring tool permissions; missing verification evidence
   - a checklist for the next real workflow, focused on: tool boundaries, state ownership, done signal, recovery path
   - one practical exercise with a clear done signal: Map one current coding workflow as a harness and mark the first missing proof signal.
6. Add a "learning transfer" section: what changes in my workflow tomorrow if I actually learned this?
7. Add a "source check" section that cites which transcript anchor supports each major takeaway.

Quality bar:
- Make this specific to "Pi Extensible Workflows: Full Guide", not a generic AI Strategy essay.
- Tie each harness element to a transcript anchor that names a tool, state boundary, permission, model behavior, or verification step.
- Prefer operational examples, failure modes, and reusable artifacts over broad definitions.
- Call out uncertainty instead of smoothing over weak evidence.
- Avoid these generic drifts: generic agent definitions; model leaderboard claims; tool list without operating boundaries.
- If evidence is weak or missing, stop and say what transcript segment or timestamp needs review instead of guessing.
- Finish with a concise artifact I could paste into my learning app.