notescraft/specs/plugins-parity/clean-paste/tasks.md

Tasks: clean paste, HTML to Markdown (roadmap 29)

One task = one agent (specs/agents/opencode-runbook.md). Spec: spec.md. Rules: AGENTS.md, docs/CODE-MAP.md. First tool call rule: after at most three reads, write the first OWNS file in your SECOND tool call (compiling stub + ONE failing test), then make it pass. Standing VERIFY: export PATH="$HOME/.cargo/bin:$PATH"; cargo test -p <pkg>; cargo clippy -p <pkg> --all-targets -- -D warnings; rustfmt --edition 2024 <owned files>; UI tasks add cargo xtask i18n; last task of a phase adds cargo xtask layers and cargo xtask tree. Standing DO NOT: edit files outside OWNS, git state, cargo fmt --all; add allow; unwrap/expect/panic/[i]/[a..b] on input-derived data; add a dependency (the crate has none).

Owner/lead step 0

Answer Q1 to Q3. Collect three real fixtures into crates/htmlmd/tests/fixtures/ (Word, Google Docs, a browser article): the owner or lead pastes text/html clipboards (macOS: osascript -e 'the clipboard as «class HTML»'); task 3 cannot finish without them.

Phase 1 - pure crate

• 1. Crate notescraft-htmlmd: tokenizer, entities, limits (brief briefs/16-htmlmd-tokenizer.md; includes crate registration). OWNS: crates/htmlmd/Cargo.toml, crates/htmlmd/src/lib.rs, crates/htmlmd/src/tokens.rs, crates/htmlmd/src/entities.rs, crates/htmlmd/tests/tokens.rs; registration: root Cargo.toml (one [workspace.dependencies] line), xtask/src/layers.rs (("htmlmd", Class::Layer(0))). FILES TO OPEN / GOAL / TESTS / VERIFY / DO NOT: see the brief.
• 2. Blocks and inlines to GFM. OWNS: crates/htmlmd/src/convert.rs, crates/htmlmd/src/inline.rs, crates/htmlmd/src/list.rs, crates/htmlmd/src/table.rs, crates/htmlmd/src/lib.rs (to_markdown, Options, Converted exports), crates/htmlmd/tests/convert.rs. FILES TO OPEN: spec.md D1, crates/htmlmd/src/tokens.rs, crates/markdown/src/lib.rs (how GFM is parsed here, for the round-trip test: use notescraft_markdown only as a dev-dependency? NO: layers forbid same-layer deps. Use pulldown-cmark already approved as a DEV dependency in crates/htmlmd/Cargo.toml and say so in the report). GOAL: D1 and the escaping rules; iterative stack, depth cap; tables with a header row and pipes escaped; code blocks fenced with a fence longer than any backtick run inside. TESTS: each construct; nested lists 4 deep; ordered list start; <br> in cell; escapes at line start; depth 65 -> TooLarge; round trip of 6 sample Markdown docs through pulldown-cmark HTML. VERIFY: standing -p notescraft-htmlmd. DO NOT: add runtime deps; handle Word quirks yet.
• 3. Word and Google Docs quirks, safety, fixtures. OWNS: crates/htmlmd/src/quirks.rs, crates/htmlmd/src/convert.rs (hook lines), crates/htmlmd/tests/fixtures/*.html + *.md, crates/htmlmd/tests/fixtures.rs, crates/htmlmd/tests/hostile.rs. FILES TO OPEN: spec.md D2 and D4, the fixtures. GOAL: D2 heuristics (inline style weight/italic/strike, MsoList levels, Docs wrapper, conditional comments, o:p); URL scheme filter. TESTS: acceptance 1, 2, 3 and 5 (LCG-seeded tag soup). VERIFY: standing. DO NOT: read files outside tests/fixtures.

Phase 2 - engine and UI

• 4. Command edit.html_to_markdown and setting. OWNS: crates/engine/Cargo.toml (one dep line), crates/engine/src/clipboard_md/{mod,commands}.rs (module name must not clash: check git grep -n "mod clipboard" crates/engine/src), one mod line in crates/engine/src/lib.rs, one register_* call in commands.rs::register_core, crates/engine/src/prefs/paste_config.rs, crates/engine/src/prefs/catalog/ entry, crates/engine/src/ai/assistant_gate.rs, crates/engine/tests/core_commands.rs, crates/engine/tests/html_to_markdown.rs. FILES TO OPEN: crates/engine/src/viewers/ (small command group to copy), crates/engine/src/prefs/files_config.rs, crates/engine/tests/files_config_viewers.rs. GOAL: command + editor.paste_html setting (convert|plain, invalid rejected). TESTS: command converts a fixture; TooLarge maps to a typed error; setting default and validation. VERIFY: standing + --test core_commands. DO NOT: touch the UI.
• 5. Read HTML from the clipboard and insert. OWNS: crates/ui-egui/src/shell/attachment_drop/clipboard.rs (Html variant + read), crates/ui-egui/src/shell/attachment_drop/mod.rs (Request::Text, routing; Shift+V = plain), crates/ui-egui/src/shell/app/editor/attachments.rs (handle Request::Text), crates/ui-egui/src/text/paste.rs + [ui.paste] keys at END, crates/ui-egui/tests/paste_html.rs. FILES TO OPEN: spec.md D3, those files, crates/ui-egui/src/testsupport.rs (fake ClipboardSource). GOAL: acceptance 3 and 4; verify arboard::Get::html() compiles (Q1: stop and report if it needs a feature). TESTS: fake clipboard with HTML -> one undo step with converted text; Shift -> plain; setting plain; TooLarge -> plain + toast; image on clipboard still follows the old path. VERIFY: standing + cargo xtask i18n. DO NOT: break image/file paste; do vault IO while painting.
• 6. Docs and status. OWNS: ROADMAP.md, docs/ARCHITECTURE.md (new crate), docs/CODE-MAP.md, specs/plugins-parity/roadmap.md (mark 29). VERIFY: cargo xtask ci.