Created: 2026-09-07 · Last-updated: 2026-09-07 · Status: DRAFT
The decisions in the chapters are intended to remain useful when the reference projects change. This appendix fixes the evidence used in this construction draft. Links point to exact commits, not a moving main branch. The stated interpretation is ours unless the cited source explicitly documents the rationale.
| Project and pin | Primary evidence | Use in this book |
|---|---|---|
OpenClaw 354538083db0a8728e16238cbd0b7a304416ff24 | Gateway architecture | Chapters 1 and 8 compare a gateway's channel/session boundary with one thin adapter. |
| OpenClaw, same pin | Session writer delivery authority | Chapter 12 compares concrete fencing scopes. OpenClaw has fencing; we make no claim that ours is unique. |
NanoClaw acc69a70962af6707aa8a6abba699bdaa7da95f8 | README | Chapter 3 examines the documented choice to delegate reasoning to the Claude Agent SDK, against our owned loop. |
| NanoClaw, same pin | Host sweep | Chapter 9 examines events as hints and a durable rescan after missed events. |
| NanoClaw, same pin | Container runner | Chapter 14 compares retaining a session with terminating a bounded report. It does not claim NanoClaw lacks recovery. |
Hermes d538f4e9297d7fa46193f638215d002d7a22edd7 | Memory tool | Chapter 5 compares prompt memory and the documented cache-prefix rationale with explicit preference records. |
| Hermes, same pin | Skills tool | Chapters 6 and 16 inspect progressive skill disclosure and shared-skill provenance; that helper does not establish the project's complete activation policy. |
Recheck a comparison before extending it
Read the linked function and its callers at the pin. Identify the problem it solves and the boundary within which its checks apply. Quote only a short documented rationale, then label any inference about trade-offs. An experiment that would change our choice is more useful than a ranking of entire projects by one feature.
Do not extrapolate module counts, vulnerability anecdotes or a single observed default into claims about a whole project's security. The feedback that motivated this edition contained assertions of absent fencing and unsupported field-wide comparisons. Source inspection corrected those claims. They are not repeated as selling points.
NemoClaw and OpenShell are candidates for a later deployment comparison, not audited reference implementations in this edition's current map. Adding them requires a distinct teaching decision and pinned primary evidence. The optional Zeocore bridge is a separately executed interoperability example, not a fourth substitute for the loop.
Keep building with Prof Rod
Found this material through a colleague, classroom or shared download? Get the complete book at profrod.ai/book and join the Prof Rod learner community. Bring one result, one question or one failure you learned from. Share this resource with another learner and keep its source links with it so they can find the full course and future updates.
