hermes-ops

Hermes instance ops: gateway, Discord, profiles, pets.

SKILL.md

GitHub

Hermes Instance Ops — this install's operational playbooks

Upstream Hermes framework knowledge lives in the bundled hermes-agent skill (CLI, config keys, contributing). This skill is the counterpart for operating this instance: the failures, rebuilds, and subsystem quirks we actually hit here, kept as verbatim case files so the next session doesn't re-derive them.

Trigger: the bot's messaging misbehaves, a gateway fights itself, a profile needs to be stood up as a new being, or a Hermes display extra (mascots) needs fixing.

Playbooks

1. Gateway / messaging troubleshooting → references/gateway-troubleshooting.md

Use when messaging is wrong in ways /new does NOT fix: DM system-message spam (tool bubbles, credit notices, memory notices, empty posts), "I can see your thinking", typing ghosts, gateway churn ("Another gateway instance is already running", "Gateway (re)started N times in 120s"), or lock fights.

Core principle to state first: /new resets the session, not config or processes — persistent weirdness is display.* or a duplicated gateway process. Heredocs of the fix commands are in the reference.

Case files (verbatim session forensics): references/2026-08-20-dm-noise-and-gateway-war.md, references/restart-war-20260820.md, references/2026-09-01-lumi-ghost-forensics.md, references/2026-09-03-lumi-unfreeze-fix.md, references/context-compression-diagnostics.md, references/multi-profile-operations.md. Scripts: scripts/gateway-forensics.sh, scripts/gw-lumi-vesper-fix.sh, scripts/discord_create_channel.py.

2. Profile identity bootstrap → references/profile-identity-bootstrap.md

Standing up a NEW Hermes profile as a distinct being (not a clone of an existing one), and adopting another AI's openly-shared lorebook frameworks with correct attribution. Covers the inherited-memory audit, depersonalizing an adopted framework, SOUL/MEMORY/USER voice-carry, and the lorebook loading mechanism.

Triggers: "make a new profile for <name>", "give Vesper their own identity", "adopt <X>'s frameworks", "clone these files but make them mine".

Support files: references/declone-workflow.md, references/depersonalize-open-framework.md, references/framework-adaptation.md, references/lorebook-audit-adapt-existing.md, references/lorebook-loading-mechanism.md, references/lorebook-refresh-workflow.md, references/shapes-login-flow.md, templates/ingest-lorebooks-openrouter.py.

3. Petdex mascots → references/petdex-mascots.md

Install / select / scale / customize the animated pet spritesheets, and debug rendering across CLI, TUI, desktop app, and SSH sessions (graphics-capable terminal requirements, config, filesystem issues). Spritesheet spec: references/spritesheet-format.md.

4. Discord server management & moderation → references/discord-server-management.md

Using and extending Hermes' built-in Discord REST tool (tools/discord_tool.py) for server management and moderation: the 15 existing actions, how core/admin actions split across the discord and discord_admin toolsets, the action-dispatch extension pattern (_ACTIONS / _ACTION_MANIFEST / _REQUIRED_PARAMS / _ACTION_403_HINT), and the RedBot coexistence topology.

Triggers: "moderate Discord", "kick/ban/timeout members", "extend discord tools", "add moderation actions". Raw endpoint reference: references/moderation-api-endpoints.md; helper script: scripts/discord_leave_guild.py.

Notes

Absorbed the former standalone skills hermes-gateway-troubleshooting, profile-identity-bootstrap, pet-management, and discord-server-management; their full bodies and support files live under references/, scripts/, and templates/ above.

Sibling umbrellas: self-management (my autonomy infra: crons, flags, backups, host updates), session-recall (session DB forensics), vector-memory-setup (Qdrant memory), local-inference (local model/voice providers).