Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Ertinox7711/SGRR-AGI-V2 --skill operating-hermes-botsgit clone --depth 1 https://github.com/Ertinox7711/SGRR-AGI-V2Wrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/ertinox7711/sgrr-agi-v2/operating-hermes-bots)<a href="https://agentmods.dev/skills/ertinox7711/sgrr-agi-v2/operating-hermes-bots"><img src="https://agentmods.dev/badge/skills/ertinox7711/sgrr-agi-v2/operating-hermes-bots/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/ertinox7711/sgrr-agi-v2/operating-hermes-bots"><img src="https://agentmods.dev/badge/skills/ertinox7711/sgrr-agi-v2/operating-hermes-bots.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00099 | $0.03115 |
| Opus 5 | $0.00049 | $0.01558 |
| Sonnet 5 | $0.00020 | $0.00623 |
| Haiku 4.5 | $0.00010 | $0.00312 |
Grade B, and why
operating-hermes-bots scanned grade B with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Asks for rootmediumPrivilege escalation
A mod that escalates privileges can change anything on the machine, not only the project.
- Changer le modèle à la place de the operator : `/model <tag>` dans `#anw` est à lui. Le bot ne change jamais le sien (hook `anw-model-lock.sh`). Jamais `chmod 444 config.yaml` : la gateway l'écrit sur `/model`. Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
8. Pré-check avant toute sonde : Ollama répond (`curl.exe -s -m 5 http://127.0.0.1:11434/api/version` puis `/api/ps` — **jamais `ollama ps`**, voir Pièges) et aucun jeu ne tient la VRAM (`nvidia-smi --query-gpu=memory.us How it starts
The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Operating Hermes bots
Principe
Le prompt système d'une session Hermes est figé à la création de la session (cache d'agent par session ; SOUL.md et la plupart des clés config ne sont pas dans la signature du cache). Donc : tout changement de config / SOUL / skills / hooks = restart de la gateway + /reset dans le canal, sinon rien ne change et on croit à tort que « ça ne marche pas ».
Deuxième principe : tout ce qui se dit et se décide dans une session Hermes se journalise (voir « Enregistrer »). Les leçons de la nuit meurent avec le chat si on ne les écrit pas.
Carte — deux bots, jamais confondre
| NeverGiveUp | Always Need Wins (ANW) | |
|---|---|---|
| Profil / HERMES_HOME | default / ~/.hermes |
anw / ~/.hermes/profiles/anw |
Unité systemd (scope --user) |
hermes-gateway |
hermes-gateway-anw |
| Canal | #<AGENT_CHANNEL> <DISCORD_ID> |
#anw <DISCORD_ID> |
| Modèle | claude-sonnet-5 (Anthropic, setup-token) | anw-38:latest (Qwen3.8-27B dense, IQ3_S) local Ollama Windows 127.0.0.1:11434/v1, zéro cloud |
| Parler au bot | ask-hermes / tell-hermes |
claude-in-chrome dans #anw (voir ci-dessous) |
| Logs | ~/.hermes/logs/ |
~/.hermes/profiles/anw/logs/{agent,gateway}.log, sessions state.db |
Détails (clés config, hooks, chiffres, rollbacks) : reference.md de ce dossier. Historique des décisions : journal.md.
Boucle de changement (dans cet ordre)
- Lire
journal.md(dernières entrées) et la section utile dereference.md: connaître l'état avant de toucher. - Backup horodaté :
cp fichier fichier.bak-YYYYMMDD-HHMMSS(cp toujours OK ; jamais rm / mv / écrasement). - Éditer via un script
.pyécrit dans le scratchpad et lancé par chemin littéral en WSL. Jamais de sed / heredoc /python -cinline souswsl -- bash -lic(métacaractères mangés). - Avant tout restart, vérifier que the operator n'est pas en plein échange :
grep "conversation turn" agent.log | tail -1sansresponse readyplus récent dansgateway.log= tour en vol ; regarder aussi#anw. Un restart coupe le tour en vol et déclenche la « session hygiene » (session compressée brutalement). Si un tour est en vol : attendre. systemctl --user restart hermes-gateway-anw, 9 s,is-active; si hooks :grep "shell hook registered" agent.log./resetdans le canal. Je le fais moi-même quand the operator a donné carte blanche sur#anwet n'est pas en train de travailler avec le bot ; sinon je lui demande de le taper./reset=/new, coûte un tour à froid (~30 s).- Preuve sans GPU que la règle est dans le prompt vivant :
python3 scripts/check_prompt.py anw "phrase distinctive"(litsessions.system_promptdansstate.db, lecture seule). - Pré-check avant toute sonde : Ollama répond (
curl.exe -s -m 5 http://127.0.0.1:11434/api/versionpuis/api/ps— jamaisollama ps, voir Pièges) et aucun jeu ne tient la VRAM (nvidia-smi --query-gpu=memory.used,memory.total --format=csv,noheader). Puis test live avec critère de réussite écrit avant l'envoi (réponse attendue, sans outil, longueur). Deux sondes réussies = signal, pas garantie (modèle probabiliste). - Mesurer, pas supposer :
agent.log→API call #N: model=… in=… out=… latency=…;gateway.log→response ready … time=…; footer Discordanw-38 · xx%= modèle réel + % de contexte (le tag y est lu du runtime, donc c'est lui la vérité, pas ma mémoire). - Enregistrer (section suivante). Rapport à the operator : résultat d'abord, chiffres avant/après, rollback en une commande.
What ships with it
5 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 76 lines · 99 tokens per session scan B 90c83670a393
operating-hermes-bots is a skill published in the GitHub repository Ertinox7711/SGRR-AGI-V2 (1 stars, last pushed 4d ago), licensed MIT. It adds 99 tokens to every session and 3,115 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it B with 2 findings (asks for root, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-09.
Other skills, from other repositories
taiyi-ui-design
A design-planning guide for describing how an application's user interface should look and behave. It produces a UI-DESIGN.md document covering layouts, components, interactions, accessibility, and error states.
taiyi-evolve
A workflow skill that compares the implemented code with the frozen design after development and testing. It records architecture changes and proposes updates to DESIGN.md, the document describing the intended system structure.
taiyi-diagram-c4
A code-scanning tool that builds C4 architecture documents from a repository. It separates facts observed in the code from conclusions inferred about the design and uses Mermaid diagrams as the source format.
retrospective-audit
Stage B of /prflow:retrospective-weekly: given a most-recent-first subset of one recurring pattern's occurrence-PR context bundles (bounded by auditbundlecap), re-derive the root cause and return one JSON object carrying a ranked findings array (one to three sub-patterns) — no edits, no worktree. Invoked as a subagent…
docs
Use when documentation generally needs to catch up with a branch before pushing or merging, covering internal developer docs, external customer-facing docs, and release notes together — "update the docs", "do a docs pass before I merge", "make sure everything's documented". Prefer this when no single documentation…
tmux-lane-orchestrator
Manage one tmux agent lane from its matching ops pane, inspect live pane state and Codex logs on cold start, and produce concise manager summaries for OpenClaw and adjacent project work.