localign
Casper RutjesTelling an AI agent "do not do X" is not a boundary. Removing X from its toolkit is.
The real control over agents sits in tool engineering, not in prompt engineering. Wrote down how we approach that at @localign.
Telling an AI agent "do not do X" is not a boundary. Removing X from its toolkit is.
The real control over agents sits in tool engineering, not in prompt engineering. Wrote down how we approach that at @localign.
