docs: add launch proof runbook

docs/ROADMAP.md

@@ -33,14 +33,22 @@ The v0.12.3 npm, GitHub, website, and official MCP Registry release gates are
 complete. The next launch milestone is durable, inspectable proof before any
 short-lived promotional channel:
 
-- [ ] Publish the CodeCartographer self-analysis case study.
-- [ ] Publish the forward-synthesis case study, including the unchecked proposal,
-      human confirmation, conflict ledger, and provenance-backed project plan.
-- [ ] Publish the pinned `sindresorhus/p-map` analysis before Product Hunt.
-- [ ] Capture the real dashboard and provenance-ledger screenshots and complete
-      the concise walkthrough or video.
-- [ ] Prepare Show HN, DevHunt, OpenAI Developer Community, and Product Hunt
-      assets only after their proof and account-readiness gates pass.
+1. [ ] Publish the CodeCartographer self-analysis case study from an immutable
+       v0.12.3 checkout using `full-with-deep-audit`, with configuration,
+       artifacts, verified claims, reproduction commands, validation results,
+       runtime, and available token/cost data.
+2. [ ] Publish the forward-synthesis case study, visibly demonstrating the
+       unchecked proposal, blocked transition, human confirmation, exact input
+       versions, conflict dispositions, complete provenance path, and final
+       project plan.
+3. [ ] Capture the real dashboard, proposal, conflict, provenance, and plan
+       visuals; export channel-ready variants and complete the 90–150 second
+       walkthrough/video.
+4. [ ] Pass the Show HN readiness checklist and launch it independently of
+       Product Hunt account readiness.
+5. [ ] Publish the pinned `sindresorhus/p-map` analysis before Product Hunt.
+6. [ ] Prepare DevHunt, OpenAI Developer Community, and Product Hunt only after
+       each channel's proof, asset, and account-readiness gate passes.
 
 The authoritative requirements, reproduction metadata, status, and channel
 sequencing live in the

docs/plans/2026-07-21-v0.12.1-discoverability-launch.md

@@ -316,12 +316,19 @@ Every case study must include:
 
 ### C1. CodeCartographer maps itself
 
-- [ ] Run an analysis pipeline against the v0.12.3 tag in an isolated workspace.
+- [ ] Resolve and record the immutable commit behind the v0.12.3 tag; analyze
+      that commit in an isolated clean checkout rather than a moving branch.
+- [ ] Run the `full-with-deep-audit` pipeline and preserve the pipeline
+      definition, workspace configuration, model, surface, and relevant options.
 - [ ] Publish architecture, contracts, defects, porting bundle, and
       reimplementation-spec artifacts or clearly explain any omitted phase.
 - [ ] Highlight three claims and trace each back to source evidence.
 - [ ] Include at least one inference and one unresolved question so the example
       demonstrates epistemic boundaries rather than only happy-path output.
+- [ ] Record exact reproduction commands, validation results, elapsed time, and
+      token/cost data when available.
+- [ ] Manually verify every highlighted claim against the pinned source and
+      record any artifact correction or editorial omission.
 - [ ] Write a concise case-study index that can be understood in under five
       minutes.
 
@@ -341,33 +348,69 @@ pass Gate 3. It must land before Product Hunt.
 ### C3. Forward synthesis
 
 - [ ] Reuse the deterministic setup from `npm run demo:synthesis` where useful.
-- [ ] Preserve the unchecked proposal artifact before confirmation.
-- [ ] Record the human's exact checked selections as a separate visible step.
-- [ ] Publish the pinned input specifications, vision, confirmed proposal,
-      merged-spec conflict ledger, and final project plan.
+- [ ] Publish the initial vision and every pinned input specification with its
+      exact selected version.
+- [ ] Preserve the initial proposal with every selection visibly unchecked.
+- [ ] Attempt the next transition before confirmation and capture the blocked
+      result so the runtime gate is demonstrated rather than merely described.
+- [ ] Record the human's exact checkbox edit and preserve the confirmed proposal
+      as a separate visible artifact.
+- [ ] Publish the merged-spec conflict ledger, including each conflict's
+      disposition, and the final provenance-backed project plan.
 - [ ] Highlight one final decision that cites the vision, one that cites a
       confirmed spec version, and one labeled synthesis inference.
+- [ ] Trace at least one complete provenance path from vision or source evidence
+      through the confirmed proposal and conflict disposition to a final plan
+      decision.
 - [ ] Show a conflict that remains visible instead of being silently flattened.
 - [ ] State that runtime and prompt contracts are tested while live-LLM plan
-      quality remains an evaluation boundary.
+      quality remains an evaluation boundary; this case study evaluates
+      generated-artifact usefulness and does not replace automated tests.
 
 ### C4. Screenshots and video
 
-- [ ] Capture a real phase dashboard and a readable project-plan provenance
-      ledger. Do not use fabricated UI.
-- [ ] Export a 1270x760 version of at least two images for Product Hunt and
-      web-optimized variants for the README/site.
+- [ ] Capture real, legible views of the phase dashboard, unchecked proposal,
+      confirmed proposal, conflict ledger, decision provenance, and final
+      work-package plan. Do not use fabricated UI.
+- [ ] Export clean 1270x760 gallery versions for Product Hunt and 3:2,
+      web-optimized variants for DevHunt and the website.
 - [ ] Add meaningful alt text and keep important text legible on mobile.
-- [ ] Produce one short demo video with this arc: unfamiliar repository →
-      evidence-backed spec → explicit confirmation → provenance-backed plan.
+- [ ] Produce a 90–150 second execution-led demo rather than a slide deck, with
+      this arc: unfamiliar repository → evidence-backed spec → blocked
+      transition → explicit confirmation → conflict handling →
+      provenance-backed plan.
 - [ ] Add captions and a text transcript.
-- [ ] Embed or link the video from README, website, case studies, and launch
-      drafts after checking the permanent URL.
+- [ ] Embed or link the video from the README, website, both core case studies,
+      and launch drafts after checking the permanent URL.
 
 **Gate 3 passes when:** the self-analysis and forward-synthesis cases are public,
-their highlighted claims have been manually verified, both required screenshots
-exist, and the video or an equivalent concise walkthrough is available. The
-small-repository case may follow before Product Hunt.
+their highlighted claims have been manually verified, the required screenshot
+set exists, and the video or an equivalent concise walkthrough is publicly
+available and embedded. The small-repository case may follow before Product
+Hunt.
+
+### Launch-ready verification
+
+Before Show HN:
+
+- [ ] Reproduce the documented demo from a clean checkout.
+- [ ] Confirm both core case studies identify their immutable source commits and
+      CodeCartographer version.
+- [ ] Confirm the published synthesis case visibly shows the unchecked proposal,
+      blocked transition, human edit, and confirmed proposal.
+- [ ] Follow at least one complete provenance path from vision or source
+      evidence to a final plan decision.
+- [ ] Open every screenshot, walkthrough/video, artifact, install, npm, GitHub,
+      website, and Registry link from a fresh/private browser session.
+- [ ] Confirm the Hacker News account can submit and reserve time for the
+      maintainer to participate after posting.
+- [ ] Finish Show HN-specific copy and first-comment notes; do not reuse one
+      generic launch post across communities.
+
+Product Hunt account eligibility, the pinned small-repository case, square
+thumbnail, and Product Hunt gallery are Product Hunt readiness requirements.
+They do not block Show HN. Each later channel must pass its own checklist in
+Workstream D rather than inheriting unrelated account or asset requirements.
 
 ## Workstream D — Prepare channel assets
 
@@ -395,6 +438,8 @@ Proposed title:
 
 - [ ] Link to GitHub so readers can inspect and run the project immediately.
 - [ ] Ensure the README offers a short path that does not require signup.
+- [ ] Pass the Show HN items in the launch-ready verification checklist; Product
+      Hunt account readiness is not a dependency.
 - [ ] The maintainer should personally write the submission text/first comment
       from the fact sheet: why it exists, what evidence-backed means, how to run
       it, why the human gate exists, and what feedback is wanted.
@@ -447,6 +492,7 @@ Proposed tagline:
 - [ ] Write the maker's first comment with the story, users, key capabilities,
       limitations, and a request for feedback—not upvotes.
 - [ ] Draft and schedule only after all three case studies are available.
+- [ ] Confirm the Product Hunt account is eligible to submit before scheduling.
 - [ ] Treat Day 7 as “prepare and schedule,” not necessarily launch day. If the
       account is new or the assets need iteration, launch later.