Skip to content

Fix canonical chain drift and homepage defaults across the repo#1

Merged
hnshah merged 1 commit into
mainfrom
claude/assess-repository-g6d9l
Apr 14, 2026
Merged

Fix canonical chain drift and homepage defaults across the repo#1
hnshah merged 1 commit into
mainfrom
claude/assess-repository-g6d9l

Conversation

@hnshah
Copy link
Copy Markdown
Owner

@hnshah hnshah commented Apr 14, 2026

Summary

  • Align the whole repo on CLAUDE.md's 6-step canonical method plus an optional claim-check durability pass.
  • Remove homepage hardcoding from every quickstart prompt, every guide prompt, and the wedge-definition template.
  • Retire the page-architecture step (framework, template, guided-run) since it is not part of the canonical chain.
  • Renumber guided-runs/ to an unambiguous 01-07 sequence (was two-03, two-04).
  • Bring claim-checking to parity: rewritten framework, new template, new guided-run 07.
  • Add frameworks for first-page-decision and page-argument-shape so the two newest canonical steps finally have durable method docs.
  • Rewrite Claude, Perplexity, Grok and OpenAI guides to match the ChatGPT guide's per-step prompt-block format, so each guide is genuinely self-contained.
  • Honesty fixes: run-logging framework now distinguishes fully logged vs summary logged; vegan-dog-food run is flagged as summary logged; README and examples README no longer overclaim what the current runs prove.
  • Remove empty examples/typeahead/* placeholder shells and the orphan templates/page-brief-template.md.

Test plan

  • Skim README.md, START-HERE.md, CLAUDE.md, and all 5 guides to confirm they describe the same 6-step chain.
  • Walk guided-runs/01 through 07 in order and confirm every Next: link resolves.
  • Grep the method docs for "homepage" and confirm every remaining mention is a warning, not a default.
  • Confirm runs/vegan-dog-food-first-pass/evaluation.md carries the summary-logged disclosure.

Align the full repo on CLAUDE.md's 6-step canonical method plus an
optional claim-check durability pass. Previously five different docs
described five different chains and four of five prompt entry points
hardcoded "homepage" into the first prompt.

Structural changes:
- renumber guided-runs so the sequence is unambiguous 01-06
- retire the page-architecture step (framework, template, guided-run)
- add frameworks for first-page-decision and page-argument-shape
- bring claim-checking to parity: framework rewrite, new template,
  new guided-run 07
- remove empty typeahead example shells
- remove orphan page-brief template

Drift fixes:
- normalize the 6-step chain across README, START-HERE, all guides,
  all quickstarts and start-with-prompts
- rewrite claude, perplexity, grok and openai guides to match the
  ChatGPT guide's per-step prompt-block format
- remove homepage hardcoding from every quickstart prompt and from
  the ChatGPT and Claude guide prompts
- remove homepage hardcoding from the wedge-definition template
- update README repo-structure to list guides, research and
  START-HERE.md

Honesty fixes:
- distinguish "fully logged" vs "summary logged" runs in the
  run-logging framework
- flag vegan-dog-food-first-pass as summary logged, not fully logged
- remove README overclaim that worked examples are "coming later"
- note in examples README what the current runs do not yet prove
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants