The problem
TODO: two or three paragraphs. What the manual job is, where it breaks, what “good” looks like. Use the vocabulary of the target query.
How it works
TODO: what it reads, what it is steered by (standards, skills), what it produces, where a human checks.
Where it gets it wrong
TODO: the known failure modes and what we do about them. Mandatory.
Why not just prompt a chatbot
TODO: three or four sentences. The steering, the skills, the guardrails and the fit to your repo and standards are the work; the model is not.
What you get
- In your engagement: TODO one line.
- Under your brand: TODO one line.
- Built to own: TODO one line.