Give an Agent Tools, Not Unrestricted Access
How I separate observation, scoped changes, secrets, and destructive operations when building a homelab agent.
Tag
Agents
6 posts filed under this topic.
Posts
From Alert to Verified Repair: Giving a Homelab Agent an Ops Job
How I traced failed NAS monitors through Uptime Kuma, SNMP, and volatile Unraid paths—and verified what the repair actually proved.
Migrating a Home DNS Stack Without Turning It Into a Network Outage
A serial, reversible way to replace a legacy home DNS resolver with AdGuard Home—using protocol checks and proof points instead of faith.
Discord Buttons Turn an Agent Into a Workflow
How native Discord buttons make agent approvals explicit, safer, and easier to automate without building a custom bot UI.
We Turned a Memory Wiki Into an Operational Knowledge Base
How an OpenClaw agent setup moved beyond passive memory by separating note types, compiling and linting the wiki, syncing it into Obsidian, and shaping pages around action for both humans and agents.
I Built a Petty Bureaucrat for My Data Broker Cleanup
How I modeled data broker removal as a stateful workflow with case tracking, inbox processing, anti-bot experiments, scheduled rechecks, and hard approval boundaries.