Set up a Hermes agent that helps a multi-team engineering organisation stay aligned, unblock work, find ownership, reduce silos, and answer common steering questions from Slack, JIRA, Confluence, and GitHub.
The repo is designed for organisations where managers, tech leads, TPMs, product partners, and engineering leadership need a shared assistant that can:
- Explain current priorities and objectives.
- Route ownership and escalation questions.
- Summarize blockers, dependencies, and decision points.
- Find architecture, repo, JIRA, and Confluence context.
- Draft clearer updates, tickets, summaries, and follow-ups.
- Support AI-native ways of working without removing human accountability.
- Escalate sensitive or ambiguous topics to the right person or forum.
Setup has two flows.
Point Codex, Claude Code, or another coding agent at this repo and ask it to set up the scaffold.
The coding agent prepares the repo, Hermes profile, source registry, permissions, initial bootstrap state, validation checks, and handoff packet. It should not ask anyone to fill context packets manually.
Use:
After install, Hermes reads the handoff packet and starts a lightweight calibration conversation. Hermes asks only the questions it cannot infer from sources, then drafts updates to the agent context.
Use:
- templates/hermes-first-conversation.md
- agent/bootstrap.md
- agent/context/09-bootstrap-state.md
- agent/context/10-leader-calibration.md
- Paste templates/coding-agent-install-prompt.md into Codex or Claude Code.
- Provide organisation name, role/context, first audience, available systems, seed links, and hard boundaries.
- Let the coding agent configure the scaffold and write handoff/hermes-handoff.md.
- Paste templates/hermes-first-conversation.md into Hermes.
- Answer Hermes's calibration questions.
- Let Hermes draft context, gaps, evals, and next actions from connected sources.
- INSTALL.md: setup instructions for coding agents.
- AGENTS.md: root instructions for coding agents and Hermes.
- hermes-profile/: Hermes SOUL, USER, MEMORY, config, and environment scaffold.
- handoff/hermes-handoff.md: bridge from coding-agent install to Hermes self-setup.
- agent/bootstrap.md: Hermes self-bootstrap loop.
- agent/soul.md: bot mission, posture, voice, and boundaries.
- agent/instructions.md: behaviour rules across Slack, JIRA, Confluence, and GitHub.
- agent/manifest.md: context loading and maintenance contract.
- agent/permissions-and-actions.md: action and approval matrix.
- agent/source-registry.yaml: machine-readable source and integration registry.
- agent/context/: agent-consumable context packets.
- agent/workflows/: discovery playbooks for Slack, JIRA, Confluence, and GitHub.
- agent/routines/: recurring Hermes routines for blockers, objectives, stale docs, PR ownership risk, manager questions, and corrections.
- templates/: install prompts, calibration prompts, evaluation templates, source inventory, and launch checklists.
- docs/runbook.html: visual overview of setup flows, context, governance, and operating cadence.
Hermes can start narrowly before everything is complete. For useful private use, prioritize:
- Source authority and freshness rules.
- Org topology and ownership routing.
- Objectives and company priorities.
- Team charters and active work.
- Architecture and codebase index.
- Operating cadences and decision forums.
- Leadership principles and escalation rules.
- AI-native transformation expectations.
- Governance, privacy, and sensitive-topic boundaries.
LICENSE
INSTALL.md
AGENTS.md
agent/
bootstrap.md
soul.md
instructions.md
manifest.md
permissions-and-actions.md
source-registry.yaml
context/
workflows/
routines/
hermes-profile/
handoff/
docs/
runbook.html
templates/
coding-agent-install-prompt.md
hermes-first-conversation.md
The bot should be useful before it is complete.
Coding agents install the rails. Hermes bootstraps the context. Humans provide judgement, verification, and approval where it matters.
MIT. See LICENSE.