33god-projects

name: 33god-projects description: > Create, wire, and maintain 33god/DeLoNET projects. Covers pjangler/CommonProject bootstrap, Hermes PM and scrum-master/Ticket Sentinel provisioning with inherited profile config, mandatory mise/.env.op wiring, BMAD, Hindsight/Bloodbank hooks, and shared Hermes fleet updates. Use when running pjangler init, pjangler hermes-agent, or mise run init-project; adding a PM or ticket sentinel; updating Hermes for the fleet; changing the fleet default model/config; touching hermes-agent-template or pjangler provisioning; wiring mise/op inject; installing BMAD; configuring agent hooks; or wiring the per-dev, committed project-scoped hook + skill fan-out so teammates and every agent CLI (Claude/Codex/Hermes/Kimi) inherit the same hooks and skills. Keywords: pjangler, CommonProject, hermes-agent-template, .project.json, Ticket Sentinel, inherited profile config, Hermes update, BMAD, Hindsight, Bloodbank, project-scoped hooks, skill fan-out, .agents/local.json, defer_to_global, hindsight-setup, hooks.master.json.

33god Project Creation & Wiring

Every 33god/DeLoNET repo is assembled by pjangler (the installed deployer CLI) out of two copier templates it vendors as submodules under ~/code/pjangler/templates/:

• CommonProject (templates/commonproject) — the base skeleton: .project.json, mise.toml, .mise/scripts/, BMAD, the ticket board.
• hermes-agent-template (templates/hermes-agent) — Hermes agent roles (PM, scrum-master/Ticket Sentinel, dev, …) provisioned into agents/hermes/<role>/.

.project.json at the repo root is the single source of truth for project + board identity. There is one ticket board per repo; every agent binds to it.

Operating Principles

• .project.json is canonical. Board binding (ticket_provider block), repo_path, project_slug, and the agents map live there. Never reintroduce a separate .plane.json.
• One board per repo. The PM owns it; the Scrum Master sentinel watches the same board. Board name = the project name (no role suffix); identifier = slug[:4] uppercased.
• Agent config is inherited by default for new fleet agents. pjangler creates ~/.hermes/profiles/<repo>-<role> as a named profile that points at the role's agents/hermes/<role>/runtime/ repo and opts config.yaml into inheriting from the fleet default profile. Local agent config.yaml files contain only overrides, such as terminal.cwd; .env, SOUL, memories, sessions, skills, gateway state, cron, and runtime files stay local.
• mise is mandatory and uniform. Every repo gets the same mise.toml contract (below).
• Agents are memory- and event-wired by default. Hindsight recall/retain + Bloodbank emit/consume are part of provisioning, not an afterthought.
• Hooks and skills fan out from the repo, per-dev. A repo that adopts the project-scoped agent layer commits one hooks SSOT + one skill set and lets mise enter/leave install them into each dev's CLIs, so teammates inherit them with zero setup (see references/project-scoped-hooks.md). Reference impl: CAF.
• Templates are version-locked. pjangler runs the vendored submodule, not an ad-hoc checkout, unless PJANGLER_HERMES_TEMPLATE overrides it for template development.

Route by intent

Read only the topic you need. Most tasks touch exactly one.

The standard lifecycle (at a glance)

1. CommonProject  →  mise run init-project        # repo skeleton + Plane board + .project.json + BMAD
2. pjangler hermes-agent (role: pm, +companion)   # PM agent, inherited profile, repo board
   └─ companion provisions the scrum-master (Ticket Sentinel) on the SAME board
3. mise trust && direnv-style `enter`             # links AGENTS.md, op-injects .env.op → .env

Steps 1–2 are detailed in references/project-creation.md; the mise enter behavior in references/mise-conventions.md.

Cross-cutting rules (apply to every 33god repo)

• AGENTS.md is the source of truth; CLAUDE.md and GEMINI.md are symlinks to it (kept current by the link-agentfiles mise task + a watch_files trigger).
• Secrets live in .env.op (1Password references); mise enter runs op inject -i .env.op > .env. Never commit .env; .env.op holds only op:// refs.
• Shared non-secret Hermes settings live in the fleet default ~/.hermes/config.yaml. New PM and scrum-master profiles inherit that config through profile.yaml: config.inherit_from: default and config.save_mode: delta. When the fleet default model changes, the repo agents follow automatically unless their local profile has an intentional override.
• No code changes in a hermes-managed repo without an active ticket on the repo board (ALLOW_NO_TICKET=1 is the emergency bypass).
• Board creation is outward-facing (creates a real Plane/Linear/Trello board) — confirm before running provisioning that hits a live workspace.

Out of scope

• Developing pjangler itself (new Commands/Recipes, the Command/Recipe pattern) → the pjangler-dev skill in the pjangler repo (~/code/pjangler/skills/pjangler-dev).
• General DeLoNET host conventions (paths, zshyzsh, Docker/Traefik, vault) → delonet-conventions.
• Hindsight API usage / bank routing mechanics (beyond the agent-hook wiring) → the hindsight skill.
• Hermes PM template governance / propagating rules to existing agents → hermes-pm-template-maintenance.
• BMAD workflow execution (PRD, stories, dev-story, sprint) → the bmad-* skills/agents.
• Migrating a stale/split-brain repo (e.g. an old PM board + separate SM board) is a deliberate manual reconciliation, not covered by the standard recipe.