I Wanted a Guardian. I Kept Building Things I Had to Babysit.
Description and use
The author built an autonomous Guardian agent that watches GitHub repositories, identifies changes, and reports findings, reducing manual oversight.
Problem
The author needed a companion agent to reduce manual oversight of multiple agents and to autonomously investigate and report on repository changes.
Outcome
The Guardian can autonomously run scheduled checks, detect and report repository changes, and update its own lessons, but its effectiveness in consistently delivering useful investigations remains to be fully validated.
Workflow
- 1. Install Hermes Agent and the SCS Guardian package. 2. Configure a GitHub watcher to emit repository snapshots. 3. Schedule a daily 7 AM job that runs the watcher and feeds the snapshot to the Guardian. 4. The Guardian processes the snapshot, identifies changes, and generates a report. 5. The Guardian updates its procedural lessons and emits a final answer to the author.
- 6. The author reviews the Guardian’s report, verifies corrections, and iterates on the watcher or Guardian logic.
- 7. Repeat the cycle, refining the Guardian’s investigation and reporting.
- 8. Use the Guardian’s lessons to improve future runs.
Evidence
The public package contains a companion template, a procedural skill drawn from our corrections, and a GitHub watcher that gives it something concrete to investigate.
The watcher is one tool. It is not the whole Guardian, and its list of observations is not a list of everything the agent is allowed to think or say.
The portable collector uses Python's standard library. It records repository revisions, pull request identities, forks, workflow files, collaborators, and Actions runs.
The biggest implementation decision was to stop writing the entire agent runtime ourselves.
The first monitor tick also invokes the agent with baseline context. There is no previous observation to call a discovery.
- Record date: 2026-10-22T22:11:29.659Z