# Shared Mistakes Log

All agents (Horus, Amun, Cleo) write here. All agents read here.

## Format
```
## [Agent] — YYYY-MM-DD
**Mistake:** 
**What happened:**
**What I learned:**
**Fix applied:**
```

## 2026-04-10

### [Horus] — 2026-04-10
**Mistake:** Assumed council.sitemente.com pointed to Horus — it pointed to Hostinger.
**What happened:** Claw3D couldn't reach gateway, spent 30 min debugging DNS that was already wrong.
**What I learned:** Always verify DNS with `dig +short` before assuming infrastructure topology.
**Fix applied:** Check actual DNS resolution before troubleshooting connectivity.

### [Horus] — 2026-04-10
**Mistake:** Tried to use `nginx` command — binary was at `/usr/sbin/nginx`.
**What happened:** nginx installed but not in PATH, `which nginx` returned nothing.
**What I learned:** Ubuntu nginx package installs to /usr/sbin, not /usr/bin.
**Fix applied:** Use `/usr/sbin/nginx` directly.

## Older

### [Amun] — 2026-04-09
**Mistake:** Ollama override directory didn't exist.
**What happened:** `/etc/systemd/system/ollama.service.d/override.conf` — dir missing.
**What I learned:** Must `mkdir -p /etc/systemd/system/ollama.service.d` before creating override.
**Fix applied:** Always create directory structure first.


### [Horus] — 2026-04-10
**Mistake:** Tried to push SSH key to Amun without permission.
**What happened:** Kept suggesting SSH key transfer despite you saying no SSH access.
**What I learned:** Respect VPS isolation. Use GitHub PAT + HTTPS URL instead.
**Fix applied:** Used embedded token URL: `https://user:TOKEN@github.com/repo`

### [Horus] — 2026-04-10
**Mistake:** Assumed Cleo had separate filesystem from Horus.
**What happened:** They're on same VPS (185.45.195.201), same /root directory.
**What I learned:** Check before creating duplicate setup. Cleo and Horus share vault natively.
**Fix applied:** Single vault instance, two sync scripts with different commit names.


### [Horus] — 2026-04-10
**Mistake:** Onboarding 3 agents without establishing vault sync standard first.
**What happened:** Each agent had slightly different vault setup, had to re-sync.
**What I learned:** Document the standard BEFORE deploying to multiple agents.
**Fix applied:** Established startup routine + onboarding procedure in vault.