Reproduction
Data Liberation Agent 0.7.3 captures https://moenadeghe.com/books and /journal. A paired Static Site Importer + Blocks Engine candidate imports 24 pages with zero unsupported fallbacks and 36/36 editable forms, but Books and Journal each retain a bare <input type="search"> as a core/html block (block_grammar); core_html_block_count=2 keeps quality red. On the live Books page, entering zzzz-no-match changes the collection to “No books match your search”; on imported WordPress the input is inert. The live Journal currently has no entries, so its effective filter behavior cannot be proven from a nonempty collection.
Producer issue Automattic/data-liberation-agent#466 tracks bounded typed-input and collection-state evidence. This consumer should use that producer contract, not convert the bare input to core/search (which submits a global WordPress query rather than filtering this page’s items) or assume a runtime predicate from the placeholder alone.
Acceptance
- Given neutral captured typed-search transition evidence linking one field to a concrete editable collection and empty-state result, generate native Gutenberg blocks or a minimal Blocks Engine-owned custom block where core blocks cannot express the verified behavior. The filter’s query and target items are owner-editable, and the source-backed response works at 390/768/1440px.
- Verify matching/no-match queries against live source and imported page; preserve category-button interactions and item visibility. Editor edit/save/reopen keeps the search control and collection valid. No
core/html island remains for a proven search field.
- For incomplete evidence (for example Journal’s empty source collection), explicitly retain the unsupported state rather than claiming functioning search without proof.
- Baseline-failing neutral regression and fresh 24-route import demonstrate the true block outcome and no behavior loss.
AI assistance
OpenAI GPT-6 Sol via OpenCode inspected DLA interaction states, live Books typing, two importer core/html emissions, and the existing core/search converter, then drafted this issue under Chris Huber’s direction.
Reproduction
Data Liberation Agent 0.7.3 captures https://moenadeghe.com/books and /journal. A paired Static Site Importer + Blocks Engine candidate imports 24 pages with zero unsupported fallbacks and 36/36 editable forms, but Books and Journal each retain a bare
<input type="search">as acore/htmlblock (block_grammar);core_html_block_count=2keeps quality red. On the live Books page, enteringzzzz-no-matchchanges the collection to “No books match your search”; on imported WordPress the input is inert. The live Journal currently has no entries, so its effective filter behavior cannot be proven from a nonempty collection.Producer issue Automattic/data-liberation-agent#466 tracks bounded typed-input and collection-state evidence. This consumer should use that producer contract, not convert the bare input to
core/search(which submits a global WordPress query rather than filtering this page’s items) or assume a runtime predicate from the placeholder alone.Acceptance
core/htmlisland remains for a proven search field.AI assistance
OpenAI GPT-6 Sol via OpenCode inspected DLA interaction states, live Books typing, two importer core/html emissions, and the existing core/search converter, then drafted this issue under Chris Huber’s direction.