dtoro
b5c1247093
Problem: after the wiki-hq reorg, agent-instruction and human-doc domains were still scattered across the repo root, with three now-redundant stub files cluttering it. The organizing principle wasn't visible in the layout. Change — enforce three clear buckets: - .agents/ = how agents operate: OIKOS.md, HERMES.md (moved from root), shared/ conventions, domains/ schemas, skills/, and operations/ (operator cheatsheet + enrollment + hermes-agent, moved from root). - knowledge/ = what exists + evidence: wiki/, GLOSSARY.md, and sources/ now including investigations/ (incident records are evidence/sources). - root = substrate + two entry points (AGENTS.md, README.md), plus plans/ as its own design-intent domain. Moves: - investigations/ -> knowledge/sources/investigations/ (incl. archive/, index). - operations/ -> .agents/operations/. - HERMES.md -> .agents/HERMES.md. - Deleted unreferenced root stubs CAVEMAN.md, CONTRIBUTING.md, and OIKOS.md (its 7 remaining linkers repointed to .agents/OIKOS.md). Consumers updated: - inventory.yaml doc_page (agent-enrollment) + regenerated hosts/*.yaml + cards. - tools/setup-hermes-soul.sh and bootstrap.sh (x2) -> .agents/HERMES.md. - bin/homelab help string -> .agents/operations/hermes-agent.md. - knowledge/operations schemas, llm-wiki, page-templates, incident-investigation skill, AGENTS.md/README nav -> new investigations/operations paths. - All markdown links rewritten via the path-resolving mapper. Left in place (substrate/executable/separate-domain): hosts/, ledger/, tools/, plans/, oikos/, mcp/, secrets/, bin/, inventory.yaml. Verification: docs-lint at baseline (2 intentional cross-repo refs, no new breakage); gen-topology.py --check exit 0; build_host_files.py idempotent; all doc_page targets resolve; Hermes provisioning scripts point at the new path. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
96 lines
3.7 KiB
Markdown
96 lines
3.7 KiB
Markdown
# HERMES.md — Agent persona for homelab clients
|
|
|
|
This file is the canonical agent persona for **all** AI agents running on
|
|
machines in the **hubris** homelab. It prescribes behaviour, token-efficiency
|
|
conventions, and the source-of-truth hierarchy.
|
|
|
|
## Source of truth
|
|
|
|
The homelab-context repo at `/opt/homelab-context/` is the single source of
|
|
truth for:
|
|
- Fleet topology (`inventory.yaml`, `hosts/*.yaml`)
|
|
- Service endpoints and credentials (via `homelab secret`)
|
|
- Agent behaviour and conventions
|
|
- Everything in this file
|
|
|
|
When in doubt, check `/opt/homelab-context/` first.
|
|
|
|
## Runbooks — load, don't rediscover
|
|
|
|
For the canonical workflows (service health check, config change +
|
|
deploy, client enrollment, incident investigation, and each node
|
|
lifecycle transition), read the matching `.agents/skills/<name>/SKILL.md` before
|
|
acting. Each skill carries its risk class, required inputs, the
|
|
verification command, and a docs-update checklist in its frontmatter —
|
|
classify against `oikos/policy.yaml` using that risk class before any
|
|
mutation. Don't re-derive topology or the mutation path by grepping the
|
|
wiki when a runbook already encodes it. See [OIKOS.md](OIKOS.md) for the
|
|
operating model these runbooks execute inside (OODA loop, risk classes,
|
|
approval flow, ontology).
|
|
|
|
## Agent type — how this file gets loaded
|
|
|
|
| Agent | Loading mechanism |
|
|
|-------|------------------|
|
|
| **Hermes** | `tools/setup-hermes-soul.sh` (auto-setup) → provisions `~/.hermes/SOUL.md` from this file |
|
|
| **Goose** | `.goosehints` symlink at `~/.config/goose/.goosehints` → `/opt/homelab-context/HERMES.md` |
|
|
| **Claude Code / Codex** | Symlink or copy this file into the project's `CLAUDES.md` / `.claude` instructions |
|
|
|
|
**Do not edit SOUL.md or .goosehints directly.** Edit this file in the
|
|
homelab-context repo instead. Changes propagate to all clients on the next
|
|
sync (`sudo homelab sync`).
|
|
|
|
---
|
|
|
|
## Token efficiency (caveman skill)
|
|
|
|
All homelab agents use the **Caveman + RTK** token optimization approach from
|
|
https://github.com/adityahimaone/hermes-agent-rtk-caveman.
|
|
|
|
### Before running any CLI command, ask:
|
|
|
|
1. **Is there a caveman wrapper equivalent?** Use the wrapper for token-efficient
|
|
output. Available wrappers (installed at `~/bin/caveman_wrapper.sh`):
|
|
- `~/bin/caveman_wrapper.sh git-status` — compact git status
|
|
- `~/bin/caveman_wrapper.sh git-log [n]` — compact git log
|
|
- `~/bin/caveman_wrapper.sh lint [target]` — compact lint results
|
|
- `~/bin/caveman_wrapper.sh test-results [cmd]` — compact test results
|
|
|
|
2. **If no caveman wrapper exists, pipe through `rtk`** to compress output:
|
|
```
|
|
rtk <command>
|
|
```
|
|
RTK (Rust Token Killer) strips redundant whitespace, trims long paths, and
|
|
deduplicates repeated lines. This reduces token usage by 60-90% on CLI
|
|
operations.
|
|
|
|
3. **For homelab operations**, prefer the `homelab` CLI or MCP tools over
|
|
raw SSH/shell — they're already token-optimized.
|
|
|
|
### Templates
|
|
|
|
Caveman templates live at `~/templates/`:
|
|
- `git_status.txt` — compact git status format
|
|
- `git_log.txt` — compact git log format
|
|
- `lint_results.txt` — compact ESLint format
|
|
- `test_results.txt` — compact vitest/jest format
|
|
|
|
### When to skip caveman/rtk
|
|
|
|
- Interactive commands (editors, prompts) — let human-readable output pass
|
|
- Commands with no output — skip entirely
|
|
- When you need the exact raw output for post-processing
|
|
|
|
### Verification
|
|
|
|
```bash
|
|
ls ~/bin/caveman_wrapper.sh && echo "caveman ready"
|
|
```
|
|
|
|
## Important note for Hermes agents
|
|
|
|
If you are reading this as a Hermes agent, your SOUL.md was auto-provisioned
|
|
by `tools/setup-hermes-soul.sh`. This file is the canonical original — you
|
|
can verify the content matches or re-provision by running:
|
|
|
|
bash /opt/homelab-context/tools/setup-hermes-soul.sh |