Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Agents · Hermes Bot Teams that don't loop (Fable 5.1)

    1

    Set up a team of specialist Hermes bots — roles, SOUL.md, per-role tool pruning, memory isolation, turn caps — with a checker that catches loops before the run.

    $9

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Agents · Hermes Bot Teams that don't loop (Fable 5.1)

    Agents · Hermes Bot Teams that don't loop (Fable 5.1)

    Example session with this skill installed

    Set up a research team in Hermes bot mode: one bot coordinates, one investigates with citations, one files approved results into my wiki. My last try looped for twenty minutes. Check the team before I run it.

    • Read your context and instructions
    • Compiled the agents hermes bot

    Summary

    Three-bot research team defined (orchestrator, researcher, librarian), souls written with the team protocol, tools pruned per role, memory isolated, runtime knobs set. The checker was run on your current team first (it fails on the classic loop makers) and on the corrected team (it passes).

    Team

    BotJobHands toNever
    orchestratorowns the task, sets limits, requests one revision, waits for you before filingresearcher, librariansearches the web itself
    researcherinvestigates with citations, saves briefs under the shared workspaceorchestratorfiles into the wiki, spawns sub-agents
    librarianfiles approved results into the wiki-edits content, researches

    Capabilities per role

    • researcher: task delegation OFF, code execution OFF, session search OFF, cron OFF; web search and browser ON
    • orchestrator: task delegation OFF, web search OFF, browser OFF; external memory ON (only writer)
    • librarian: wiki write ON (only writer); built-in memory only
    • keep-warm 5, idle timeout 10 min, group room with a turn cap of 6

    Checker, current team (team_draft.json)

    $ python3 scripts/bot_team_check.py team_draft.json
    WARN  researcher: researcher has cron/scheduling ON - automation belongs to the orchestrator
    WARN  keep_warm=1 is below the number of bots (3) - cold bots slow every mention
    WARN  idle_timeout_min=2 is short - long tasks will stall when a bot falls asleep
    FAIL  orchestrator: orchestrator has web search ON - it must delegate research, not do it
    FAIL  researcher: researcher has task delegation ON (the loop maker) - switch it off
    FAIL  researcher: researcher can write to the wiki - only the librarian writes
    FAIL  librarian: soul is empty or too short to define a job
    FAIL  more than one external-memory writer (orchestrator, researcher) - collisions; keep it on the orchestrator only
    FAIL  more than one wiki writer (researcher, librarian) - one writer to the library
    FAIL  task delegation ON for: orchestrator, researcher - the team IS the delegation; switch it off
    FAIL  group_room.turn_cap missing or not a positive integer - no loop protection
    RESULT: FAIL (8 violation(s), 3 warning(s))
    

    Checker, corrected team (team.json)

    $ python3 scripts/bot_team_check.py team.json
    RESULT: PASS (0 warning(s))
    

    Next steps

    • Restart the gateway so the edited souls load, then give the orchestrator one bounded task: "state of the art for multi-agent workflows, ten sources, under five hundred words".
    • Watch the activity view: who was tagged, who declined, when the turn settled. The librarian files nothing until you approve.
    • Afterwards ask all bots to update their memory from the session and fold what you learned into the souls and the tool matrix.
    Checker, current team (team_draft.json

    Connects securely to your tools. The creator never sees your data.

    About this skill

    The problem

    A bot in Hermes bot mode is a profile with its own SOUL.md, model, tools, skills and memory. Left at defaults, a team of them loops: workers delegate to sub-agents, two bots write the same external memory, bot-to-bot chat has no natural end, and a twenty-minute run burns tokens without a result.

    What it does

    • Defines one job per bot (orchestrator, researcher, librarian or any specialist) with an outcome, owner, deliverable and stop condition written into each SOUL.md.
    • Prunes tools and skills per role from a tested matrix; task delegation is switched off on every bot, the orchestrator included.
    • Isolates memory: built-in memory on every bot, the external memory provider on the orchestrator only.
    • Sets keep-warm and idle timeout so bots stay awake for research-length runs.
    • Picks the communication mode: direct mention, turn-capped group room, or kanban with human approvals.
    • Ships bot_team_check.py, a deterministic checker that fails loudly on the known loop and memory mistakes before you start a run (exit 0 or 1, --selftest).

    Frameworks & tools

    Hermes Agent desktop app with bot mode enabled; Python 3 standard library for the checker. No network access.

    Why this beats prompting it yourself

    Measured on the same task in two separate sessions: without the skill the delivered team failed the anti-loop checker on 2 rules and warned on 2 more (delegation left on, two memory writers); with the skill it passed 0/0. The tool matrix, the team protocol and the turn-cap rule are the parts a generic prompt does not know.

    Use cases

    • A research team: one bot coordinates, one investigates with citations, one files approved results into a wiki.
    • A coding squad with separate implementation and review roles.
    • A content pipeline where writer and editor coordinate without looping.

    Known limitations

    Does not install Hermes (see the DevOps · Install and Harden listing) and does not cover non-Hermes frameworks. Note for reviewers: the checker is standard-library Python, reads one local JSON file and makes no network access.

    Requires Hermes Agent desktop app with bot mode enabled. The configuration checker requires a local Python 3 environment. Does not automate Hermes core installation.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 2 months ago

    Requires

    Python

    Permissions

    Terminal / Shell
    Read Files

    The bundled checker (scripts/bot_team_check.py) is standard-library Python: it reads one local JSON team file passed on the command line, prints violations, exits 0 or 1. No network access, no environment variables read, no files written.

    Listed2 months ago

    What's inside

    Frequently Asked Questions

    This skill provides a pre-configured architecture and a deterministic Python checker to prevent infinite loops, redundant memory writing, and token-wasting delegation cycles common in multi-bot setups.

    It is designed specifically for the Hermes Agent desktop app with bot mode enabled; it does not support non-Hermes frameworks or standard LLM chat interfaces.

    You receive a specialized tool-pruning matrix, optimized SOUL.md templates for specific roles, team communication protocols, and the bot_team_check.py validation script.

    The included checker script requires the Python 3 standard library, but it operates entirely locally without network access to ensure your team configurations remain private.

    While the skill provides the orchestration framework, you must have the Hermes Agent desktop app already installed (refer to the 'DevOps · Install and Harden' listing if you need setup help).

    Yes, the skill is designed to work with any specialist roles you define, provided they follow the outcome, owner, and deliverable structure outlined in the provided SOUL.md templates.

    New in Agents & Orchestration