Back to search
hermes-multi-agent-workflow
SkillGitHub
A reusable skeleton for an autonomous, multi-agent triage pipeline built on Hermes: a fleet of agents that detects items from sources, dedups them, scores them against a rubric, researches them in…
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/tonbistudio--hermes-multi-agent-workflow)HTML
<a href="/listings/tonbistudio--hermes-multi-agent-workflow"><img src="/api/listings/tonbistudio--hermes-multi-agent-workflow/badge.svg" alt="findsafeskills safety rating" /></a>Install
i
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
pip install -r requirements.txt # just PyYAML python -m cli.triage validate # check the example config python -m unittest discover -s tests # engine + board-client tests, no live Hermes needed python -m cli.triage scaffold # print the Hermes setup plan
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 115 lines
View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)
- # Hermes Multi-Agent Workflow
- A reusable skeleton for an **autonomous, multi-agent triage pipeline** built on
- [Hermes](https://github.com/NousResearch/hermes-agent): a fleet of agents that
- **detects** items from sources, **dedups** them, **scores** them against a rubric,
- **researches** them in parallel, **routes** each to a fulfillment path, pauses at
- **one human approval gate**, then **fulfills and delivers** — all coordinated on a
- single Hermes Kanban board.
- It ships pre-wired as a worked example (find pain points AI-agent users hit →
- build a fix or make an explainer video), so you can read a complete pipeline and
- then repoint it at your own domain.
- > **This is a template, not a turnkey app.** It runs its unit tests and validates
- > its config out of the box, but going live requires setting up your Hermes
- > install, profiles, auth, and scouts (see `docs/07-runbook.md`). The point is to
- > give you — and your coding agent — a clear, working structure to adapt.
- ## The idea
- ```
- sources → intake → dedup → score → research (parallel) → route
- │
- ┌──────────────────────┬──────────────────────┤
- path A path B shelve
- (prep) (prep) (auto)
- └──────────┬───────────┘
- ── HUMAN GATE ── approve · shelve · modify
- ┌──────────┴───────────┐
- fulfill fulfill
- └──────────┬───────────┘
- deliver
- ```
- The shape is fixed; **what flows through it is yours.** Everything domain-specific
- lives in one file, `triage.yaml`.
- ## Quickstart
- ```bash
- pip install -r requirements.txt # just PyYAML
- python -m cli.triage validate # check the example config
- python -m unittest discover -s tests # engine + board-client tests, no live Hermes needed
- python -m cli.triage scaffold # print the Hermes setup plan
- ```
- ## Adapt it to your domain
- The whole adaptation is editing `triage.yaml` + the markdown templates it points
- at. Hand your coding agent **`AGENTS.md`** and ask it to walk you through
- `docs/04-adapting-to-your-domain.md`. In brief:
- 1. Edit `triage.yaml`: sources, rubric, research lanes, route map, paths, roles.
- 2. Edit `paths/` templates (scope rails, deliverable specs, proposal formats).
- 3. Edit `skills/templates/` (scout queries + orchestrator notes).
- 4. `python -m cli.triage validate`, keep `tests/` green.
- 5. Follow `docs/07-runbook.md` to set up profiles and go live.
- ## Repository layout
- ```
- triage.yaml THE config — your whole pipeline (start here)
- AGENTS.md Guide for the AI agent adapting this template
- engine/ Generic engine (rarely edited)
- config.py Loads + validates triage.yaml
- engine.py TriageEngine — all deterministic step logic
- scoring.py Rubric scoring (LLM mode + deterministic mode)
- routing.py Classification → path
- dedup.py Similarity (token-cosine; embedding-ready)
- item_vault.py One markdown file per tracked item
- hermes_board.py Writes the Hermes Kanban board (via `hermes kanban`)
- intake_parser.py Parses scout reports
- frontmatter.py Stdlib YAML-frontmatter for item files
- proposal_actions.py Human-gate handler (approve/shelve/modify) — config-driven
- paths/ Per-path templates you customize
- rails/ specs/ proposals/
- skills/templates/ Scout + orchestrator SKILL.md templates
- cli/triage.py validate / scaffold / init / install
- scripts/cost_report.py Per-item spend for the cost gate
- tests/ Generic engine tests
- docs/ Deep-dive docs (architecture, board, config, adapting, …)
- examples/ Reference configs
- ```
- ## Documentation
- - `docs/01-architecture.md` — fat engine / thin skill; how the pieces fit.
- - `docs/02-the-board.md` — Kanban as the bus; dispatcher; fan-in.
- - `docs/03-config-reference.md` — every `triage.yaml` key.
- - `docs/04-adapting-to-your-domain.md` — the step-by-step adaptation guide.
- - `docs/05-pipeline-stages.md` — each stage, and the gotchas to preserve.
- - `docs/06-security.md` — trust surface, scope rails, safe publishing.
- - `docs/07-runbook.md` — profiles, board, crons, go-live.
- - `examples/ai-agent-pain-points/REFERENCE.md` — full write-up of the reference
- implementation this template was extracted from.
- ## Security
- This template runs LLM-authored code and shells out, behind one human gate. Read
- **`SECURITY.md`** and `docs/06-security.md` before deploying — and run the
- pre-publish secret-scan checklist before open-sourcing an adapted copy.
- ## Contributing
- See **`CONTRIBUTING.md`**. The golden rule: keep `engine/` domain-agnostic; new
- domains go in `triage.yaml`, not the code.
- ## License
- MIT — see `LICENSE`.
- ## Credits
- Extracted and generalized from a working single-machine Hermes pipeline. The
- engine is domain-agnostic; the bundled example reflects its origin.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.