GitHub Identity Switch to n0ctilucent and OpenClaw Agent Directory Integrity
Github Identity Switch to n0ctilucent
Franklin directed OpenClaw to stop using the devsecfranklin GitHub account and adopt the n0ctilucent account as the official identity. The n0ctilucent account (user ID 87308972, company "pale shadow") was authenticated via a provided PAT:
- Token verified via API: login =
n0ctilucent, user ID 87308972 - Full API access confirmed (no scope restrictions detected)
- Profile README created at
n0ctilucent/n0ctilucentfollowing the devsecfranklin pattern — professional bio with ORCiD link, education at CU Denver, current role at Palo Alto Networks, Twitch-linked content.
Configuration Attempt and Correction
The initial attempt to set tools.github directly in openclaw.json via a raw PAT failed validation:
profileIdwas truncated (ghp_UA...Qsea) and failed the pattern check^ghp_[a-f0-9]{32}$kind: "pat"is invalid — only"oauth"is allowed for managed GitHub identity in OpenClaw.
The config auto-restored from backup. The correct path forward uses the gateway action secrets:set github_oauth_code with a device code from https://github.com/login/device, which triggers the OAuth flow and stores credentials securely in OpenClaw's managed secrets store.
OpenClaw Agent Directory Integrity — Assessment
The ~/.openclaw/agents.entries file was missing entirely on chonk, causing OpenClaw to emit warnings:
"Found 1 agent directory on disk without a matching agents.entries entry."
This means config-driven routing, identity, and model selection would ignore all agent directories. Sessions and auth state still existed on disk but config-driven features were broken.
Agent Directories Found on Disk (20 total)
| Directory | Purpose |
|---|---|
| main | Default autonomous orchestration agent |
| infra | Lab infrastructure management |
| author | Creative writing and content generation |
| puzzle | Puzzle-solving and CTF analysis |
| research | Research documentation and analysis |
| website | Website maintenance and publishing |
| teaching-ismg2050 | Course page generation and management |
| panw-steward | Palo Alto firewall policy auditing |
| database-agent | Database administration tasks |
| mc-manager | Minecraft server management |
| media | Media generation and processing |
| genealogy | Family history documentation |
| jetson | NVIDIA Jetson device management |
| code-reviewer | Code review and static analysis |
| finance | Financial tracking and analysis |
| discovery | Infrastructure discovery and reconnaissance |
| openclaw | OpenClaw configuration management |
| an | Additional needs (unspecified) |
| archive | Archived and retired agent entries |
Status
Franklin was presented with the scope of what would be needed to fix (audit vs. auto-generate agents.entries). No action was taken this session — pending user decision on which approach to take.
OpenClaw Full Backup Plan — Design
A tiered backup strategy has been designed for the full OpenClaw workspace (~/.openclaw/), targeting /mnt/clusterfs/openclaw/{date}/:
Tiered Structure
- Critical tier: openclaw.json, .env (credentials), agent JSONs (
agents/*/agent.json), all workspaces (workspace*/*MEMORY.mdetc.), agents.entries - Important tier: memory/, extensions/, plugins/, skills/, skill-workshop/
- Nice-to-have tier: discord/, devices/, credentials/, crestodian/, panw-steward/
Exclusions (never backed up)
- cache/, completions/, logs/, sandboxed/sandboxes/ — regenerable or ephemeral
- .openclaw-*.lock files
- .git/ directories within workspaces (tracked content doesn't need backup)
Retention Policy
Keep last 3 dated copies, prune older automatically.
CURRENT BACKUP STATE (Pre-Rotation)
| Type | Date | Size/Notes |
|---|---|---|
| Single file backup | 2026-10-01 11:04 | openclaw.json.bak |
| Manual backup | 2026-09-29 23:01 | openclaw.json.manual.bak |
| Tar.gz archive | 2026-09-27 10:32 | 1.5 GB (oldest) |
| Tar.gz archive | 2026-09-30 11:22 | 2.4 GB (newest tar.gz, stale for full workspace) |
Gap identified: no backup of agent workspaces, memory files, extensions, plugins, or .env exists yet on a dated basis. All existing backups cover only openclaw.json.
Key Questions for Franklin
- Which approach as primary? Tar.gz (full point-in-time) or tiered rsync (granular restore)?
- Should the clusterfs git repo be part of rotation strategy, or just agent metadata tracking?
- Retention confirmed as 3 copies — applied to both tar.gz and dated directories?