Essential session storage is required to sign in. Optional analytics remain off unless you allow them.
Pick a real, repeatable task (e.g. summarize support tickets, classify reviews by sentiment, or extract a few fields). Write a v1 prompt, run it on ~10 representative inputs, and record where it fails. Then iterate — changing ONE thing at a time (clarity, an added constraint, a few-shot example, a format directive) — to v3 or beyond, keeping a log of each change and why. Deliver the final prompt plus the before/after results showing improved reliability. Use delimiters to separate instructions from input.
The final prompt is explicit about task, audience, length, and format; input is delimited; ambiguity removed.
A real change log: one variable at a time, driven by observed failures on actual inputs — not guesswork.
Before/after on the same ~10 inputs shows the final prompt is more reliable; failures characterized.
Few-shot / constraints / format directives used where they actually help, not cargo-culted.
Use this README structure to make the work reviewable.
DevPath fetches a read-only, size-capped snapshot. Submitted code is never executed.
# A prompt iteration log ## Goal Pick a real, repeatable task (e.g. summarize support tickets, classify reviews by sentiment, or extract a few fields). Write a v1 prompt, run it on ~10 representative inputs, and record where it fails. Then iterate — changing ONE thing at a time (clarity, an added constraint, a few-shot example, a format directive) — to v3 or beyond, keeping a log of each change and why. Deliver the final prompt plus the before/after results showing improved reliability. Use delimiters to separate instructions from input. ## Acceptance criteria - [ ] Clarity & specificity: The final prompt is explicit about task, audience, length, and format; input is delimited; ambiguity removed. - [ ] Iteration discipline: A real change log: one variable at a time, driven by observed failures on actual inputs — not guesswork. - [ ] Measured improvement: Before/after on the same ~10 inputs shows the final prompt is more reliable; failures characterized. - [ ] Technique fit: Few-shot / constraints / format directives used where they actually help, not cargo-culted. ## Verification - [ ] Document installation and run commands. - [ ] Record automated checks and their exact commands. - [ ] Add representative output, screenshots, or a short demo where useful. - [ ] Confirm failure paths and known constraints. ## Decision log ### Decision title - Context: - Choice: - Alternatives considered: - Tradeoffs: ## Evidence - Link each rubric criterion to the file, test, or artifact that demonstrates it. ## Known limitations - List what is intentionally out of scope and what should be improved next.