Saw your note about safety TODOs and the rough edges around autonomous delegation — I've been building claude-guard which handles exactly this at the Claude API level, validating and filtering what gets sent to worker agents before they run rather than relying on the model to self-police. Might fit cleanly into the Worker pipeline you've got. Happy to share how it's structured if useful.
Saw your note about safety TODOs and the rough edges around autonomous delegation — I've been building claude-guard which handles exactly this at the Claude API level, validating and filtering what gets sent to worker agents before they run rather than relying on the model to self-police. Might fit cleanly into the Worker pipeline you've got. Happy to share how it's structured if useful.