# SKLD — Project Journal ## Entry #13: The Install Test — three quality gates, still broken **Date**: April 11, 2026 **Session Duration**: ~12 hours across two long stretches **Participants**: Matt + Claude Opus 4.6 (1M context) --- ### The Starting Point Entry #12 closed with the SKLD-bench content layer shipped — 867 Elixir challenges across 7 lighthouse families, all on main. This session picked up from there with a specific goal that kept expanding: take the phoenix-liveview mock pipeline run (shipped in PR #18 the previous session) and turn the production Registry page for that run into something people would actually want to click through. What started as "polish the run detail page" ended up being: 1. A ground-up rebuild of `/runs/:runId` into a 7-tab rich showcase 2. A full `mock → seed` rebrand across every user-facing string 3. A second rebrand of the run_id itself because "mock" was still visible in the URL bar 4. OG meta tags + a brand image + per-run server-side meta injection so links would render rich previews when dropped in Discord 5. And then — finally, near the end — the install test that discovered the whole thing was silently broken --- ### Phase 1: The rich run detail page The starting state: the Registry page for the phoenix-liveview mock run loaded, but it was a "skill with no story". Raw SKILL.md preview, synthetic Fitness Radar fed hardcoded data, empty Growth Curve, 12 variant rows hidden behind a "Show Advanced" toggle, export buttons at the bottom. Nothing explained what the 12 evolved capabilities did, how the composite was assembled, which challenges it was tested against, or why the composite scored 0.94. A first-time visitor would bounce. Matt framed the problem plainly: "it's a skill with no story." The fix was a complete restructure. Over several iterations with real-time feedback: - **7 tabs, sticky header**: Composite, Competition, Metrics, Tests, Narrative, Lineage, Package. Sticky tab bar with the always-visible header showing run title, Gen 1/1 pill, export buttons. - **Plain-English first, metrics second**: an `OverallAssessment` prose card at the top of the Composite tab, followed by a `PipelineOverview` mini-diagram (24 challenges → 12 variants → 1 composite), then the rendered composite SKILL.md with section anchors. - **Per-challenge competition breakdown**: `CompetitionBracket` with 12 mini-brackets, each showing Variant 1 (seed) vs Variant 2 (spawn) with per-challenge scores and a `buildRationale()` sentence explaining why the winner won. Preempts questions like "did you run baselines?" and "is this multi-gen?" - **Master-detail lineage**: the first cut had 12 parent cards that expanded in-place below a grid, but clicking parent #7 scrolled past the click target. Fixed by moving to a sticky left rail that stays visible while the right panel updates on selection. - **Stacked parent→composite sections**: side-by-side comparisons squeezed both into narrow columns and made reading hard. Matt: "let's put these on top of each other instead side by side." One line of CSS. - **Package Explorer**: split installable (SKILL.md + scripts/ + references/ + assets/ + test_fixtures/) from metadata (PACKAGE.md, REPORT.md, parents/, challenges/). SKILL.md at top, selected by default, directories open by default. A Gold Standard Checklist above the tree showing one-line factual compliance with the §CLAUDE.md quality bar. Every iteration Matt flagged another thing: "there's nothing to view here" (the expanded card was below the fold), "shouldn't there be directories here?" (the installable tree was flat), "delete this disclaimer" (I'd written a paragraph explaining how the package was built; the user wanted it gone, not rewritten). Each piece of feedback codified another design principle: no disclaimers, honest tone, plain- English before metrics, production parity with what the real engine will emit. --- ### Phase 2: Rich package enrichment Halfway through the run detail work, Matt pivoted: "ok I want to make this package RICH... we're going through all of these motions, so the next mock run we do is 100% on target and when we run this in product it generates everything we need. This isn't meant to be a fake package, but an actual package we can distribute." The composite at that point had a SKILL.md string and nothing else — empty `supporting_files`. The Package Explorer was rendering "1 file" with a disclaimer about it being a post-facto reconstruction. That was not going to fly. The fix was an enrichment pass: dispatch 5 focused Opus subagents to produce each missing file (validate.sh, main_helper.py, guide.md, cheatsheet.md, anti-patterns.md, 2× starter templates, migration checklist), plus copy 6 curated test_fixtures from the family directory. The `enrich_package.py` helper merges everything into the composite genome's `supporting_files` dict via a direct SQL UPDATE, then the seed JSON gets re-exported. End result: 16 files, ~167KB uncompressed. Gold Standard Checklist all green. Visual evidence of a real distributable package. --- ### Phase 3: Rebrand mock → seed (twice) Matt: "no one cares whether or not it actually ran through the product pipeline or not." The visitor cares about the output, not the provenance. Every user-facing "mock" became "seed". Directory renames, log prefix updates, meta_strategy values (`mock_pipeline_winner` → `seed_pipeline_winner`), script renames, loader module updates, hash marker format (`[mock_v...]` → `[seed_v...]`). Along the way I found a latent bug: the mock_run_loader appended a fresh `[mock_v]` marker to the specialization on every reload without stripping old markers. After 3 reloads the subtitle read `[mock_v...] [mock_v...] [seed_v...]`. Fixed with a regex-strip pass before appending. Phoenix-liveview's run_id was frozen at `-mock-v1` in the first rebrand to preserve DB continuity, which felt like the right trade-off until Matt sent me a screenshot of the browser URL bar showing `https://skld.run/runs/elixir-phoenix-liveview-mock-v1`. "damn the URL says mock in it." Second rebrand: rename the run_id itself. Surgical find-and-replace across the 650KB seed JSON: 63 structural occurrences of `elixir-phoenix-liveview-mock-v1` → `elixir-phoenix-liveview-seed-v1`, 4 composite genome ID replacements, plus a `LEGACY_RUN_RENAMES` map in the loader that deletes the legacy row on boot so Railway's persistent volume gets auto-migrated. Also hit a save-order FK bug in the loader (variant_evolutions.challenge_id → challenges.id) that was latent because earlier exports had NULL challenge_ids. --- ### Phase 4: OG images + per-run meta injection With the URL finally clean, Matt wanted to drop the link in an Elixir Discord channel. Before this session, `frontend/index.html` had zero OG meta tags — any shared link fell back to bare page title. The quick-win was site-wide static tags + a 1200×630 brand PNG generated with Pillow via `uv run --with pillow python < wrapped in <%= %> — invalid HEEx <%= <.link patch={~p"/users/new"}>New user, class: "btn" %> # The `class: "btn"` was dangling instead of becoming a component attr # push_navigate worked but Routes.user_path inside it wasn't rewritten {:noreply, push_navigate(socket, to: Routes.user_path(socket, :index))} # live_redirect with an Elixir expression (not a quoted string) was skipped <%= live_redirect user.name, to: ~p"/users/#{user}" %> # :for directive landed on