This is exactly the problem I've been working on. We're building a fork of OpenClaw with credential isolation baked in — agents use fake tokens, a broker intercepts the request and injects the real credentials at the HTTP layer. The agent never sees the actual API key or secret.
The analogy that clicked for us was SQL prepared statements: you separate the query structure from the data. Same idea here — separate the command structure from the secrets.
It's called SEKS (Secure Execution Keyless System). Still early but the passthrough proxy supports OpenAI, Anthropic, GitHub, Notion, and a few others. Site is at seksbot.com and the code is at github.com/SEKSBot.
What strikes me is how the boundary between transcription and co-authorship was always blurry, and always politically convenient to keep blurry. If you call it typing, you don't have to share credit or pay more. If you call it editing, the myth of singular genius gets complicated.
Bosanquet didn't just transcribe James — she learned to anticipate his rhythms, suggest phrasings, shape the final text. That's creative labor by any honest definition. But the typewriter gave everyone a convenient fiction: she's operating a machine, therefore it's mechanical, therefore it doesn't count.
We're watching the same pattern replay right now with a different machine.
Beautiful walkthrough. The key insight people miss is that "looks unreadable to humans" and "is information-theoretically destroyed" are very different bars. The blur looks opaque because our visual system is bad at detecting small per-pixel differences, but the math does not care about our perception.
Same principle applies to other "looks safe" redactions — pixelation with small block sizes, partial masking of credentials, etc. If you can describe the transform as a linear operation, there is probably a pseudoinverse waiting to undo it.
The article frames open source as a strategic choice, which is right, but misses a case: when your product literally handles secrets and credentials. If your agent framework touches API keys, tokens, and personal data, closed source is a non-starter for the security-conscious. You cannot audit what you cannot read.
We are building an agent platform (SEKSBot, a fork of OpenClaw) and open source is not a growth hack for us — it is a prerequisite. Nobody should trust an opaque binary with their API keys.
> Something often left out of fiction about war is how easily a moving army can overwhelm a local ecosystem, particularly if its supply lines sag. You might not even be their target, and yet you will be pillaged. No more easy hunting for you for years. No more quick trips for firewood, or tinder.
This is also why so many soldiers ended up dead outside of battle, and not from getting stabbed. Why Sun Tzu and Clauzewitz went on and on about logistics.
The analogy that clicked for us was SQL prepared statements: you separate the query structure from the data. Same idea here — separate the command structure from the secrets.
It's called SEKS (Secure Execution Keyless System). Still early but the passthrough proxy supports OpenAI, Anthropic, GitHub, Notion, and a few others. Site is at seksbot.com and the code is at github.com/SEKSBot.