Two new files plus pointers from README.md and agent.md. playbook.md — companion to agent.md. Where agent.md governs what evidence is admissible as you add to the spec, playbook.md covers how to navigate the work itself: triage checklist for wire-format bugs, common debugging anti-patterns (the stale-sibling-binary trap, trusting LLM training data on Reticulum specifics, chasing intermittent symptoms with retries), the three layers of test trustworthiness, and how to work productively in a code-as-spec domain. Includes an incident registry seeded with the §6.2/§6.6 signed_data signalling bug surfaced in mobile-app today plus older HEADER_2, REQUEST path_hash, DEST_LINK, and stale-binary incidents. Append-only — every future interop fix gets a registry entry per §8. templates/AGENTS.md — drop-in boilerplate for new Reticulum implementation projects in any language. Uppercase plural matches the emerging AGENTS.md convention (Claude Code, Codex, Cursor, Copilot Workspace). Sections: read-these-first reading list, cardinal rules summary, project-specific FILL-IN placeholders, contributing-findings-back obligation, attribution. Project-specific bits use HTML comments so they're obvious to edit. §5 attribution points back to this repo and is mandatory per CC BY 4.0. templates/README.md — names the template, says where to put it, restates the attribution expectation. agent.md and README.md updated with pointers to the new files so anyone reading the front door of the repo finds them. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
1.3 KiB
templates/
Drop-in files for new Reticulum implementation projects. Copy into the root of your project and edit the marked sections.
| File | Where to put it in your project | What it does |
|---|---|---|
AGENTS.md |
Project root, as AGENTS.md |
Tells AI agents (Claude Code, Codex, Cursor, etc.) and human contributors which docs to read first, the cardinal rules for working on a Reticulum stack, and how to feed findings back upstream. Includes placeholder sections for project-specific build/architecture notes. |
Why this exists
Every alternative Reticulum implementation (Kotlin, Swift, Rust, Go, C, JS) re-discovers the same handful of wire-format gotchas (HEADER_2 conversion, LRRTT, signalling in LRPROOF signed_data, etc.). Most of those gotchas are now in SPEC.md and playbook.md. The AGENTS.md template makes it trivial to point new projects at those documents — and at the incident registry — before they spend hours rediscovering known bugs.
Attribution
These templates are CC BY 4.0. Use them freely. Keeping the attribution section pointing back to this repo is the entire ask — it's how downstream agents find their way back to the source of truth, and how the spec stays maintained.