Evidence and attribution ledger

The project narrative is first-person editorial copy directed by Maggie. The original work, decision to skip Context Layer, experience of rework, and comparison with native Codex context support are her account. Repository and session evidence corroborate the specific facts below. No causal benchmark or numeric performance gain is asserted.

Claim	Evidence	Boundary
Original portfolio engineering	p00_02 aaaf0bf (2021-11-23); 821377a (2022-02-04); 54cb623 (2022-02-13); .eleventy.js at 7915693 (2024-09-14)	Git proves implementations and dates, not exclusive authorship of every line. Maggie supplies original-work attribution. No Lighthouse or speed gain invented.
Custom reusable site composition	Historical `_includes/components/`, layouts, shortcodes, navigation and image config	Do not confuse case-study text about Citi components with the portfolio's own components.
AI transition	d85d3a2 (2025-07-25), Cursor Agent author, Maggie co-author	First reviewed explicit AI commit, not proven first AI usage.
Expert skill	codex-skills 011f9f9 (2026-05-08); current supplied SKILL.md	Versioned by this date; exact initial creation date unknown. Skill was not created in the October 11 chat.
Early lightweight decision	p00_02 ecea8a4 AGENTS.md	Exact policy: “Do not add the full Product OS scaffold unless Maggie explicitly asks for the broader lifecycle/evidence/project-dashboard system. The current preference is lightweight documentation.” Small-project rationale is Maggie's October 11 retrospective.
Shadow surfaces and stale testing target	`DOCS/NEXT_SESSION.md` historical handoff at ce69410; `DOCS/portfolio-redesign.md`	Historical playground embedded old home.html; current homepage/review work later addressed some duplication before the scaffold. Do not claim all ambiguity remained unchanged until October 11.
File noise	Selected chat lines 8290–8331	20,044 untracked = 19,973 dependencies + 12 generated files + 59 other files. Count is source-session observation, not current untracked state or 20,044 lost documents.
Missing Context Layer discovered	Selected chat beginning 8403	Lightweight docs existed; missing full lifecycle/evidence/dashboard setup, not zero context or zero documentation.
Installation	p00_02 ce69410; selected chat ends 8551	Five project records; preserve original study paths; review board and bug intake chosen by Maggie.
Preservation	`DOCS/evidence/active/context-layer-migration/preserved-files.json`	690 pre-migration paths with SHA-256: 689 Git blobs match byte-for-byte; one CSV matches after restoring CRLF line endings, and its current worktree bytes match the original manifest. All 690 contents preserved.
Duplicate packaging removed	`removed-handoff-zips.json` at ce69410	One 8,731,025-byte Frill ZIP, authorized in the source chat. No cleanup performed by this intake.
Modern model setting	source turn_context lines 8297, 8335, 8403, 8437	gpt-6-astra / high, verified directly. Native product capability growth is Maggie's account, not independently researched here.

Follow-up evidence that would strengthen the claim

Track whether later sessions resume the correct project and canonical surface; whether accepted decisions need to be re-explained; whether artifacts land under an owning record; and whether an experiment graduates without a duplicate implementation. Record concrete before/after examples. Do not assign saved hours or reduced-rework percentages without measurements.

Source scope and privacy

Only the October 11 06:48:55–07:01:05 EDT episode of the specified chat was archived. The intake deliberately excludes the rest of that thread. Selected event order and tool evidence are retained privately with explicit setup/reasoning/path redactions. No client-confidential detail pages or credentials are included as public media. The separate intake request supplies the retrospective narrative; it is not misrepresented as a quotation from the source chat.