repo gps for ai agents — spoolcast dev-log #5
A devlog about why AI coding agents hallucinate workflows when a repo does not clearly route the task.
AI hallucination is not always about fake facts. Sometimes it is the AI filling in missing routing information.
This devlog traces a real Spoolcast production problem: agents inventing menu options, ignoring existing prompts, applying the wrong pipeline, and getting confused by rule debt. The fix was not more scattered docs. It was a mechanical first step that tells the agent where it is, which workflow owns the task, and what the next allowed move is.
The broader lesson: if an AI coding agent keeps making up process, the repo may not have a clear enough front door. First locate the task. Then follow the right route. If no route exists, define one before production.
AI disclosure: this video was made with Spoolcast, the AI video system I am building. I did not manually edit it in a traditional timeline; the system helped produce the structure, script, visuals, narration, render, thumbnail, and shipping files from source material, direction, feedback, and approvals.
How it's made
Built on the Spoolcast illustration/chunk Remotion pipeline with generated scene images, Google TTS narration, rendered widescreen output, narration-only SRT captions, and a final YouTube thumbnail.