Hi Addy,
Your agent-skills repo and writings on "Process over Prose" + human orchestration have been very influential.
I wanted to share a practical, low-resource case study that strongly aligns with the philosophy:
Built Oodleforge (OodleLZ + AES precision archive tool) in 34 days solo on a 2010 Intel Core 2 Quad Q6600 (4) @ 2.400GHz, Linux Mint + Wine.
Developed a strict Human-AI Supervision Method ("Team of Two") to survive context poisoning, logical loops, and hardware limits.
Full document here: https://github.com/johna124/Oodleforge/blob/main/ia-supervision-method.md
Key alignments with your approach:
Human owns architecture + validation
Small isolated Task Contracts
Early real integration + heavy stress testing
Mechanical guardrails instead of huge prompt constitutions
Would love any feedback or thoughts on how this fits into the broader agent-skills ecosystem. Happy to answer questions.
Thanks for the excellent work on agent-skills!
Best,
johna124
Hi Addy,
Your agent-skills repo and writings on "Process over Prose" + human orchestration have been very influential.
I wanted to share a practical, low-resource case study that strongly aligns with the philosophy:
Built Oodleforge (OodleLZ + AES precision archive tool) in 34 days solo on a 2010 Intel Core 2 Quad Q6600 (4) @ 2.400GHz, Linux Mint + Wine.
Developed a strict Human-AI Supervision Method ("Team of Two") to survive context poisoning, logical loops, and hardware limits.
Full document here: https://github.com/johna124/Oodleforge/blob/main/ia-supervision-method.md
Key alignments with your approach:
Human owns architecture + validation
Small isolated Task Contracts
Early real integration + heavy stress testing
Mechanical guardrails instead of huge prompt constitutions
Would love any feedback or thoughts on how this fits into the broader agent-skills ecosystem. Happy to answer questions.
Thanks for the excellent work on agent-skills!
Best,
johna124