agentleFS
jbellsolutionsjbellsolutions/hermes-super-agent/.claude/skills/agent-os/SKILL.md

agent-os

Skill4 starsChanged 7 months agoMITView on GitHub

Drive the conversational launch of agent-os, OR orient yourself in an existing repo. Detects which mode based on whether the repo has been bootstrapped.

  • Pipes a download into a shell
  • Reads credentials

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. 2867203296/hermes-agent★ 0same name
    2. christiankoertel-sketch/adam-skills★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository7 other files it gives its agents.

The file

Show its rules and text

We split this skill into the rules it gives an agent

19 rules. About 0 are common, 5 are distinctive, and 14 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 14

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Run this check first:

    line 8
  2. Distinctive:

    Before repo bootstrapping, check whether Hermes Agent itself is installed:

    line 23
  3. Distinctive:

    If Hermes is missing, install it first with the official installer:

    line 31
  4. Distinctive:

    Ask the user, in one short message, the minimum setup questions:

    line 41
  5. Very specific:

    Use AskUserQuestion if you have it.

    line 54
  6. Very specific:

    If they picked Slack: SLACK_BOT_TOKEN, SLACK_APP_TOKEN, SLACK_SIGNING_SECRET.

    line 66
  7. Very specific:

    If Telegram: TELEGRAM_BOT_TOKEN.

    line 68
  8. Very specific:

    If web voice: LIVEKIT_* triple, plus voice provider key (OPENAI_API_KEY or GEMINI_API_KEY).

    line 70
  9. Very specific:

    If Enterprise and deployment discovery is requested: RAILWAY_TOKEN and/or DIGITALOCEAN_ACCESS_TOKEN.

    line 72
  10. Very specific:

    Use the Edit tool to fill in each value the user provided.

    line 85
  11. Very specific:

    After each, summarize the result for the user in one line.

    line 97
  12. Very specific:

    If they want to proceed with the Hermes wiring NOW, offer to drop into Stage 2 — that's a focused 1–2 hour session where you wire …

    line 107
  13. Very specific:

    Run agent-os manifest to refresh the system graph.

    line 134
  14. Distinctive:

    Read ARCHITECTURE.md for the orchestrator/runtime model.

    line 136
  15. Very specific:

    Read ETHOS.md for the unbreakable rules (don't edit vendor/, don't add a 15th framework).

    line 137
  16. Very specific:

    Use /route --tags ... to confirm which runtime your work would dispatch to.

    line 139
  17. Very specific:

    If the work spans packages, write a short plan in vault/decisions/<date>-<topic>.md first.

    line 145
  18. Very specific:

    Single-state guarantee: Every channel writes through vault_memory — no per-channel state.

    line 155
  19. Very specific:

    Update setup instructions with every update: If you add a tool, runtime, tier, deployment path, or workflow, update README/LAUNCH/setup …

    line 158

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Mode detection

Run this check first:

test -f .env && test -d .venv && echo "bootstrapped" || echo "fresh"
  • fresh → run Launch Mode (below).
  • bootstrapped → run Operating Mode (further down).

Launch Mode (fresh repo, drop-the-link flow)

The user just dropped this repo's URL into your session. You are the conversational installer. Your job: walk them through LAUNCH.md Path A and docs/hermes-first-install-walkthrough.md in plain English. Don't dump documentation at them — converse.

Before repo bootstrapping, check whether Hermes Agent itself is installed:

command -v hermes || true
hermes --version || true
hermes doctor || true

If Hermes is missing, install it first with the official installer:

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

Then run or guide hermes setup before continuing the Super Agent repo setup.

Step 1 — Confirm scope

Ask the user, in one short message, the minimum setup questions:

  1. Which commercial tier? Operator / Pro Operator / Enterprise.
  2. What business/project/offer is this for? Include first 3 workflows the agent should own.
  3. Which channels do they want active first? Slack? Telegram? Web text? Web voice? Any combination.
  4. Which model setup should Hermes use? Default suggestion: connect both GPT-5.5 and Claude Opus 4.6/4.7 for architecture/debug/security/auth/unit-test/deployment-code review, plus a cheaper worker model like DeepSeek when available.
  5. Where do they want to deploy? Local-only / Railway / DigitalOcean VPS / Docker Compose / Fly.io.
  6. Where is the shared brain? Obsidian vault path and Notion integration/database IDs, if available.

Example phrasing:

"I'll walk you through Super Agent setup like a commercial onboarding — should take 15–30 minutes depending on keys. Quick setup questions: (1) tier: Operator, Pro Operator, or Enterprise? (2) what business/project is this for and what first 3 workflows should the agent own? (3) channels: Slack, Telegram, web chat, voice, or a mix? (4) model setup: GPT-5.5 + Claude Opus 4.6/4.7 together for architecture/debug/security/auth/tests/deployment logic, plus cheaper workers? (5) deploy target: local, Railway, DigitalOcean/VPS, Docker Compose, or Fly? (6) Obsidian vault path and Notion workspace/database map?"

Use AskUserQuestion if you have it; otherwise plain prose.

Step 2 — Collect keys

Based on their channel choices, ask for ONLY the keys actually needed. Don't ask for everything in .env.example if they only picked Slack. Reference .env.example for the canonical list.

Required regardless of channel: at least one model/provider key. Preferred commercial setup is both GPT-5.5-capable OpenAI/OpenRouter access and Claude Opus 4.6/4.7-capable Anthropic/OpenRouter access, plus an optional cheaper worker model such as DeepSeek.

Shared brain fields: OBSIDIAN_VAULT_PATH, NOTION_API_KEY, and Notion database IDs for conversations, actions, decisions, agents, companies/offers, deployments, approvals, and costs when available. Notion is mandatory for the final operating system, but if credentials are not available during setup, mark Notion sync as pending rather than blocking the whole install.

Business setup fields: SUPER_AGENT_TIER, BUSINESS_NAME, BUSINESS_TYPE, FIRST_WORKFLOWS, HUMAN_APPROVAL_REQUIRED.

If they picked Slack: SLACK_BOT_TOKEN, SLACK_APP_TOKEN, SLACK_SIGNING_SECRET. Link them to https://api.slack.com/apps if they don't have an app yet.

If Telegram: TELEGRAM_BOT_TOKEN. Link to @BotFather.

If web voice: LIVEKIT_* triple, plus voice provider key (OPENAI_API_KEY or GEMINI_API_KEY).

If Enterprise and deployment discovery is requested: RAILWAY_TOKEN and/or DIGITALOCEAN_ACCESS_TOKEN. If the user asks about Orgo/managed cloud computers, explain it is optional and only collect ORGO_API_KEY when the customer/workspace needs an isolated visible desktop.

For each key the user doesn't have, ASK if they want to skip it (and disable that channel/runtime for now) or pause to grab it. Do not block.

Step 3 — Stage .env

Once keys are collected, write them to .env:

cp .env.example .env
# then edit with the values gathered above

Use the Edit tool to fill in each value the user provided. Leave the rest blank (they're optional).

Step 4 — Bootstrap

Run, in order:

./scripts/bootstrap.sh        # installs uv if missing, syncs Python, installs pnpm
uv run pytest -q              # 23 tests should pass
uv run agent-os manifest      # builds the system graph; writes vault/graph/system.yaml

After each, summarize the result for the user in one line. If something fails, debug interactively — most likely a missing dep or a key that didn't make it into .env.

Step 5 — Boot Hermes

uv run agent-os boot

This is currently a stub (Stage 2 of docs/EXECUTION-PLAN.md). It returns a JSON describing the boot intent. Tell the user this honestly. The skeleton is here; the real Hermes wiring is the next session's job.

If they want to proceed with the Hermes wiring NOW, offer to drop into Stage 2 — that's a focused 1–2 hour session where you wire vendor/hermes-agent into src/agent_os/orchestrator/boot.py for real.

Step 6 — Verify single-state (when channels are live)

Once at least two channels are wired (post Stage 7 of the execution plan), run:

uv run pytest tests/integration/test_single_state_smoke.py

Confirm the test passes. This is the load-bearing acceptance criterion for the accessibility layer.

Step 7 — Hand-off

Give the user a 5-line summary:

  • What's running (channels, model, deploy target).
  • Tier/business/workflows and approval rules.
  • What's stubbed and which stage of docs/EXECUTION-PLAN.md lights it up.
  • What the daily loops do (upgrader at 02:00, agi-audit at 02:30, heartbeat every 5 min).
  • The three operating skills they have now (/agent-os, /explain, /route).
  • A single concrete next move ("ship Stage 2 next session, then Stage 3 after").

Operating Mode (bootstrapped repo, day-to-day)

Run agent-os manifest to refresh the system graph. Then:

  1. Read ARCHITECTURE.md for the orchestrator/runtime model.
  2. Read ETHOS.md for the unbreakable rules (don't edit vendor/, don't add a 15th framework).
  3. Identify which package the work belongs to (src/agent_os/orchestrator, src/agent_os/runtimes/<name>, src/agent_os/manifest, src/agent_os/quality, src/agent_os/upgrader, src/agent_os/channels, packages/webapp, packages/dashboard).
  4. Use /route --tags ... to confirm which runtime your work would dispatch to.

For commercial/multi-business decisions, read docs/commercial-packaging.md and docs/portfolio-agent-architecture.md before recommending another Hermes instance or a shared memory layout.

For Railway/DigitalOcean discovery, follow runbooks/deployment-access.md: read-only inventory first, no mutations without explicit approval.

If the work spans packages, write a short plan in vault/decisions/<date>-<topic>.md first.

For introspection ("what's running?", "who wrote this output?", "how does X connect to Y?"), use /explain.


Hard rules (apply in both modes)

  • Never edit vendor/. It breaks the upgrader. Open an upstream PR or wrap in src/agent_os/runtimes/.
  • Never start another framework wrapper. New ideas go into a runtime adapter, a Hermes skill, or upstream.
  • Single-state guarantee. Every channel writes through vault_memory — no per-channel state.
  • Smoke tests are non-negotiable for the upgrader. A bad upstream commit silently promoted is the failure mode that takes the system down.
  • Default to Hermes. Specialist runtimes are exceptions, not defaults.
  • Update setup instructions with every update. If you add a tool, runtime, tier, deployment path, or workflow, update README/LAUNCH/setup docs in the same change.
  • Separate serious businesses. One primary Hermes can orchestrate the portfolio, but project/customer agents need isolated vaults, secrets, deploy targets, and skills.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.