html-collab 0.1.0

package/package.json

@@ -0,0 +1,61 @@
+{
+  "name": "html-collab",
+  "version": "0.1.0",
+  "description": "Turn any single-file HTML report into a portable, reviewable document. Comments, suggested edits, and deterministic merge — all stored inside the HTML file itself.",
+  "type": "module",
+  "engines": {
+    "node": ">=18"
+  },
+  "bin": {
+    "html-collab": "dist/cli.js"
+  },
+  "scripts": {
+    "build": "bun build src/cli.ts --target=node --outfile=dist/cli.js && chmod +x dist/cli.js",
+    "build:examples": "bun run build:sample && bun run build:tour",
+    "build:sample": "bun examples/build-sample.ts",
+    "build:tour": "bun examples/build-tour.ts",
+    "prepack": "bun run build",
+    "prepublishOnly": "bun run verify",
+    "record:demo": "node examples/record-demo.mjs",
+    "test": "bun test",
+    "typecheck": "tsc --noEmit",
+    "verify": "bun run typecheck && bun test && bun run build && bun run build:examples"
+  },
+  "devDependencies": {
+    "@types/node": "^22.15.0",
+    "bun-types": "^1.3.0",
+    "playwright": "^1.60.0",
+    "typescript": "^5.8.3"
+  },
+  "files": [
+    "dist",
+    "README.md",
+    "LICENSE",
+    "skill"
+  ],
+  "keywords": [
+    "html",
+    "review",
+    "comments",
+    "annotations",
+    "collaboration",
+    "local-first",
+    "single-file",
+    "merge",
+    "crdt",
+    "ai-generated"
+  ],
+  "author": "Glen Maisey <glen.maisey@gmail.com>",
+  "license": "MIT",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/glendigity/html-collab.git"
+  },
+  "publishConfig": {
+    "access": "public"
+  },
+  "homepage": "https://glendigity.github.io/html-collab/examples/tour.review.html",
+  "bugs": {
+    "url": "https://github.com/glendigity/html-collab/issues"
+  }
+}

package/skill/SKILL.md

@@ -0,0 +1,70 @@
+---
+name: html-collab
+description: Process review comments from portable html-collab files. Use when a user provides or mentions a `.review.html` file produced by `html-collab`, asks to summarize review feedback, revise the underlying report, merge reviewed copies, extract an agent brief, or unwrap final HTML.
+---
+
+# html-collab
+
+Use the `html-collab` CLI as the source of truth for parsing, merging,
+extracting, wrapping, and unwrapping reviewable HTML files. Do not scrape the
+browser UI or reimplement the embedded operation-log format.
+
+## Core Commands
+
+```sh
+html-collab wrap report.html --out report.review.html
+html-collab merge glen.review.html maya.review.html --out merged.review.html
+html-collab extract merged.review.html --format agent --out agent-plan.md
+html-collab extract merged.review.html --format markdown --out review-brief.md
+html-collab extract merged.review.html --format json --out review-bundle.json
+html-collab skill --out html-collab.SKILL.md
+html-collab unwrap merged.review.html --out report.final.html
+html-collab unwrap merged.review.html --apply-edits --out report.final.html
+```
+
+## Workflow
+
+1. If the user gives reviewed copies from multiple reviewers, run `merge` first.
+2. Run `extract --format agent` when the user wants you to revise or act on
+   the review.
+3. Run `extract --format json` when you need deterministic thread/message data.
+4. Run `extract --format markdown` for a human-readable brief.
+5. Summarize unresolved threads by section or anchor.
+6. Summarize suggested edits separately from comments, including edit IDs,
+   status, replacement text, and whether they are accepted, rejected, or deleted.
+7. Separate direct edits from questions that need author judgement.
+8. If asked to revise the report, unwrap or locate the source HTML, apply edits
+   to the clean report, then wrap the revised report for another review round.
+9. When the review cycle is finished, run `unwrap --apply-edits` if accepted
+   tracked edits should be applied to the final HTML, otherwise run `unwrap`
+   to recover the original clean HTML.
+
+## Semantics
+
+Suggested edits are deterministic instructions. Open or accepted suggested
+edits can be applied directly when the user asks for revision. Rejected or
+deleted edits should not be applied unless the user explicitly asks.
+
+Comments are reviewer intent. Treat comments as questions unless the anchor and
+message make a simple edit obvious. For example, a comment anchored to `1` that
+says `Make this 5` can be treated as an inferred replacement from `1` to `5`,
+but report it as inferred and preserve the thread ID.
+
+Resolved comments should not drive new edits unless the user explicitly asks to
+revisit resolved feedback. Deleted comments/replies should be ignored except
+when auditing review history.
+
+## Response Shape
+
+When summarizing review feedback, report:
+
+- Open thread count and resolved thread count.
+- Suggested edit counts by status.
+- Reviewers.
+- Thread IDs, edit IDs, and any review-file links from the extracted brief.
+- Direct revisions the agent can make.
+- Author decisions that block revision.
+- Any unanchored or invalid operations from the extracted bundle.
+
+Keep comments as review state. Do not silently transform comments into document
+edits unless the user asks you to revise the underlying report.