qubicl-cli 0.1.0-dev.0 → 0.1.0

package/dist/assets/computer/node_modules/playwright-core/package.json

@@ -0,0 +1,34 @@
+{
+  "name": "playwright-core",
+  "version": "1.62.1",
+  "description": "A high-level API to automate web browsers",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/microsoft/playwright.git"
+  },
+  "homepage": "https://playwright.dev",
+  "engines": {
+    "node": ">=20"
+  },
+  "author": {
+    "name": "Microsoft Corporation"
+  },
+  "license": "Apache-2.0",
+  "exports": {
+    ".": {
+      "types": "./index.d.ts",
+      "import": "./index.mjs",
+      "require": "./index.js",
+      "default": "./index.js"
+    },
+    "./package.json": "./package.json",
+    "./lib/bootstrap": "./lib/bootstrap.js",
+    "./lib/coreBundle": "./lib/coreBundle.js",
+    "./lib/utilsBundle": "./lib/utilsBundle.js",
+    "./lib/tools/cli-client/program": "./lib/tools/cli-client/program.js"
+  },
+  "bin": {
+    "playwright-core": "cli.js"
+  },
+  "types": "types/types.d.ts"
+}

package/dist/assets/computer/skills/core/docx/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Nous Research
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

package/dist/assets/computer/skills/core/docx/SKILL.md

@@ -0,0 +1,21 @@
+---
+name: docx
+description: Create, read, edit, template, validate, and review Word DOCX documents using Qubicl's pinned local document tools. Use for professional Word documents, tracked revisions, comments, structured edits, and DOCX templates.
+---
+
+# DOCX workflows
+
+Keep inputs and outputs under `/home/qubicl`. Obtain this skill's `resourceRoot` from `skill_view` and run bundled helpers with `/opt/qubicl/skills-venv/bin/python`. Qubicl computer and workstation images contain the pinned libraries; never install dependencies as part of the workflow.
+
+Useful helpers beneath `<resourceRoot>/scripts/`:
+
+- `docx_read.py FILE --text|--structure|--styles|--revisions`
+- `docx_create.py SPEC.json OUTPUT.docx`
+- `docx_edit.py` for text, paragraph, style, and table-cell edits
+- `docx_template.py TEMPLATE.docx VALUES.json OUTPUT.docx`
+- `docx_comments.py` and `docx_revisions.py`
+- `docx_validate.py FILE`
+
+Read `references/revisions-and-comments.md` only for tracked-change or comment work.
+
+Inspect before editing, write to a new output unless in-place work was requested, and validate after every mutation. On workstation computers, use LibreOffice for a final PDF render when visual layout matters; on computer computers, report that structured DOCX operations work but office rendering may be unavailable. Preserve unknown OOXML parts whenever a helper supports doing so.

package/dist/assets/computer/skills/core/docx/references/revisions-and-comments.md

@@ -0,0 +1,88 @@
+# Revisions and Comments — XML details
+
+Deep reference for `docx_revisions.py` and `docx_comments.py`. Read this
+when you need to reason about the raw WordprocessingML, extend the
+scripts, or debug an unusual document. Everyday use only needs SKILL.md.
+
+## Tracked changes (w:ins / w:del)
+
+Word records run-level tracked changes as wrapper elements inside a
+paragraph (`w:p`), in the `w` namespace
+`http://schemas.openxmlformats.org/wordprocessingml/2006/main`:
+
+```xml
+<w:p>
+  <w:r><w:t>Base </w:t></w:r>
+  <w:ins w:id="1" w:author="Editor" w:date="2026-01-02T03:04:05Z">
+    <w:r><w:t>inserted text</w:t></w:r>
+  </w:ins>
+  <w:del w:id="2" w:author="Editor" w:date="2026-01-02T03:04:05Z">
+    <w:r><w:delText>deleted text</w:delText></w:r>
+  </w:del>
+</w:p>
+```
+
+Key facts the script relies on:
+
+- Deleted text lives in `w:delText`, not `w:t` — that is why plain text
+  extraction naturally shows the "accepted" view (insertions visible,
+  deletions hidden).
+- Resolution semantics:
+  - accept `w:ins` → unwrap (move child runs up, drop the wrapper)
+  - reject `w:ins` → remove the wrapper and its contents
+  - accept `w:del` → remove the wrapper and its contents
+  - reject `w:del` → rename each `w:delText` to `w:t`, then unwrap
+- Revisions can appear anywhere block content is allowed: body, table
+  cells (nested tables too), headers, footers, text boxes. The script
+  iterates the body root plus every header/footer part root with
+  `root.iter(W+"ins", W+"del")`, which finds them at any depth.
+- `w:id` values are unique per revision *element*, but one logical edit
+  session may produce several elements. `accept`/`reject --id` acts on
+  exactly the element(s) carrying that id.
+
+Not handled by the script (detected by `docx_read.py --revisions` but
+left alone): paragraph-mark revisions (`w:rPr/w:ins` on `w:pPr`), table
+row insertions/deletions (`w:trPr/w:ins`), format-change records
+(`w:rPrChange`, `w:pPrChange`), and moves (`w:moveFrom`/`w:moveTo`).
+Moves are rare from typical editors; if present, treat the file with
+Word itself rather than guessing.
+
+## Comments
+
+Three cooperating pieces:
+
+1. **`word/comments.xml`** — one `w:comment` element per comment,
+   carrying `w:id`, `w:author`, `w:initials`, `w:date`, and body
+   paragraphs. Related from document.xml via the relationship type
+   `.../comments` and content type
+   `application/vnd...wordprocessingml.comments+xml` (also needs a
+   `[Content_Types].xml` override — python-docx's part machinery adds it
+   when the part is registered).
+2. **Range markers in the story** — `w:commentRangeStart w:id="N"`
+   before the anchored runs, `w:commentRangeEnd w:id="N"` after them.
+3. **The reference run** — a `w:r` containing `w:commentReference
+   w:id="N"`, placed right after the range end; it ties the balloon to
+   the location.
+
+`docx_comments.py` behavior:
+
+- **list / delete** always work at the XML level, so they handle files
+  from any producer. `anchored_text` is reconstructed by walking each
+  part root in document order and collecting `w:t` text between the
+  start and end markers for each id.
+- **add** first isolates the target text into whole runs. If the match
+  starts or ends mid-run, the run is split at the boundary (the split
+  copies `w:rPr`, so formatting is preserved). Then:
+  - python-docx >= 1.2: the native `document.add_comment(runs, ...)`
+    API is used (it creates the comments part, markers, and reference
+    run itself).
+  - older versions or `--xml`: the script builds `word/comments.xml`,
+    registers the part + relationship through the opc layer, and
+    inserts the markers/reference manually.
+- Deleting a comment removes the `w:comment` element and all three
+  marker kinds for that id; the anchored document text is untouched.
+
+Modern Word also writes `commentsExtended.xml` (threading/resolved
+state). The scripts neither read nor produce it: replies and "resolved"
+flags are invisible here, and comments added by this skill are plain
+top-level comments.

package/dist/assets/computer/skills/core/docx/scripts/docx_comments.py

@@ -0,0 +1,289 @@
+#!/usr/bin/env python3
+# MIT-licensed helper adapted for Qubicl's native DOCX skill.
+"""List, add, and delete comments in a .docx.
+
+Subcommands:
+  list    JSON per comment: id, author, initials, date, text, anchored_text
+  add     add a comment anchored to the first occurrence of --target
+  delete  remove a comment (and its range markers) by --id
+
+Examples:
+  docx_comments.py list report.docx
+  docx_comments.py add report.docx --target "Q3 revenue" \
+      --text "Needs a source" --author "Reviewer" -o out.docx
+  docx_comments.py delete report.docx --id 0 -o out.docx
+
+Uses the native python-docx comments API (>= 1.2) when available; falls
+back to building word/comments.xml and the range markers directly for
+older versions (or when --xml is passed). Listing and deletion always
+work at the XML level so they handle documents from any producer.
+"""
+from __future__ import annotations
+
+import argparse
+import datetime as _dt
+import json
+import sys
+from copy import deepcopy
+
+from docx import Document
+from docx.opc.constants import RELATIONSHIP_TYPE as RT
+from lxml import etree
+
+from docx_common import iter_part_roots
+
+W = "http://schemas.openxmlformats.org/wordprocessingml/2006/main"
+COMMENTS_CT = ("application/vnd.openxmlformats-officedocument"
+               ".wordprocessingml.comments+xml")
+
+
+def q(tag: str) -> str:
+    return f"{{{W}}}{tag}"
+
+
+# ---------------------------------------------------------------- reading
+
+def _comments_root(doc):
+    """Return the XML root of the comments part, or None."""
+    for rel in doc.part.rels.values():
+        if rel.reltype == RT.COMMENTS:
+            part = rel.target_part
+            el = getattr(part, "_element", None)
+            if el is not None:
+                return el
+            return etree.fromstring(part.blob)
+    return None
+
+
+def _anchored_texts(doc) -> dict:
+    """Map comment id -> document text between its range markers."""
+    anchored: dict[str, list[str]] = {}
+    for root in iter_part_roots(doc):
+        active: set[str] = set()
+        for el in root.iter():
+            if el.tag == q("commentRangeStart"):
+                cid = el.get(q("id"))
+                active.add(cid)
+                anchored.setdefault(cid, [])
+            elif el.tag == q("commentRangeEnd"):
+                active.discard(el.get(q("id")))
+            elif el.tag == q("t") and active:
+                for cid in active:
+                    anchored[cid].append(el.text or "")
+    return {cid: "".join(parts) for cid, parts in anchored.items()}
+
+
+def list_comments(doc) -> list:
+    root = _comments_root(doc)
+    if root is None:
+        return []
+    anchored = _anchored_texts(doc)
+    out = []
+    for c in root.iter(q("comment")):
+        cid = c.get(q("id"))
+        text = "\n".join(
+            "".join(t.text or "" for t in p.iter(q("t")))
+            for p in c.iter(q("p")))
+        out.append({"id": cid, "author": c.get(q("author")),
+                    "initials": c.get(q("initials")),
+                    "date": c.get(q("date")), "text": text,
+                    "anchored_text": anchored.get(cid, "")})
+    return out
+
+
+# ---------------------------------------------------------------- anchoring
+
+def _split_run(para, run_el, offset: int):
+    """Split a run element at text offset; return the new right-hand run."""
+    text = "".join(t.text or "" for t in run_el.iter(q("t")))
+    right = deepcopy(run_el)
+    run_el.addnext(right)
+    for el, s in ((run_el, text[:offset]), (right, text[offset:])):
+        for t in list(el.iter(q("t"))):
+            el.remove(t)
+        t = etree.SubElement(el, q("t"))
+        t.text = s
+        t.set("{http://www.w3.org/XML/1998/namespace}space", "preserve")
+    return right
+
+
+def find_anchor_runs(doc, target: str):
+    """Isolate `target`'s first occurrence into whole runs; return them."""
+    from docx_common import iter_all_paragraphs
+    for para in iter_all_paragraphs(doc):
+        full = para.text
+        start = full.find(target)
+        if start < 0:
+            continue
+        end = start + len(target)
+        pos = 0
+        covered = []
+        for run_el in para._p.iter(q("r")):
+            rtext = "".join(t.text or "" for t in run_el.iter(q("t")))
+            r_start, r_end = pos, pos + len(rtext)
+            pos = r_end
+            if r_end <= start or r_start >= end:
+                continue
+            if r_start < start:  # split off the left part
+                run_el = _split_run(para, run_el, start - r_start)
+                r_start = start
+            if r_end > end:      # split off the right part
+                _split_run(para, run_el, end - r_start)
+            covered.append(run_el)
+        return para, covered
+    return None, []
+
+
+# ---------------------------------------------------------------- adding
+
+def _next_id(doc) -> int:
+    root = _comments_root(doc)
+    if root is None:
+        return 0
+    ids = [int(c.get(q("id"), "0")) for c in root.iter(q("comment"))
+           if c.get(q("id"), "").isdigit()]
+    return max(ids) + 1 if ids else 0
+
+
+def add_comment_native(doc, runs, text, author, initials):
+    from docx.text.run import Run
+    run_objs = [Run(r, None) for r in runs]
+    comment = doc.add_comment(run_objs, text=text, author=author,
+                              initials=initials or "")
+    return str(comment.comment_id)
+
+
+def add_comment_xml(doc, runs, text, author, initials) -> str:
+    cid = str(_next_id(doc))
+    root = _comments_root(doc)
+    if root is None:
+        root = etree.fromstring(
+            f'<w:comments xmlns:w="{W}"/>'.encode("utf-8"))
+        from docx.opc.packuri import PackURI
+        from docx.opc.part import Part
+        blob = etree.tostring(root, xml_declaration=True,
+                              encoding="UTF-8", standalone=True)
+        part = Part(PackURI("/word/comments.xml"), COMMENTS_CT, blob,
+                    doc.part.package)
+        doc.part.relate_to(part, RT.COMMENTS)
+        # keep a live element on the part so edits reach save()
+        part._element = root
+        part.blob_ = None
+
+        def _blob(self=part):
+            return etree.tostring(self._element, xml_declaration=True,
+                                  encoding="UTF-8", standalone=True)
+        part.__class__ = type("CommentsXmlPart", (Part,),
+                              {"blob": property(lambda self: _blob(self))})
+    now = _dt.datetime.now(_dt.timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
+    comment = etree.SubElement(root, q("comment"))
+    comment.set(q("id"), cid)
+    comment.set(q("author"), author)
+    if initials:
+        comment.set(q("initials"), initials)
+    comment.set(q("date"), now)
+    p = etree.SubElement(comment, q("p"))
+    r = etree.SubElement(p, q("r"))
+    t = etree.SubElement(r, q("t"))
+    t.text = text
+    # range markers around the anchor runs + reference run after them
+    first, last = runs[0], runs[-1]
+    start = first.makeelement(q("commentRangeStart"), {q("id"): cid})
+    first.addprevious(start)
+    end = last.makeelement(q("commentRangeEnd"), {q("id"): cid})
+    last.addnext(end)
+    ref_run = last.makeelement(q("r"), {})
+    ref = etree.SubElement(ref_run, q("commentReference"))
+    ref.set(q("id"), cid)
+    end.addnext(ref_run)
+    return cid
+
+
+# ---------------------------------------------------------------- deleting
+
+def delete_comment(doc, cid: str) -> bool:
+    root = _comments_root(doc)
+    found = False
+    if root is not None:
+        for c in list(root.iter(q("comment"))):
+            if c.get(q("id")) == cid:
+                c.getparent().remove(c)
+                found = True
+    for part_root in iter_part_roots(doc):
+        for tag in ("commentRangeStart", "commentRangeEnd",
+                    "commentReference"):
+            for el in list(part_root.iter(q(tag))):
+                if el.get(q("id")) == cid:
+                    parent = el.getparent()
+                    # remove the wrapping run for reference marks
+                    if tag == "commentReference" and parent.tag == q("r"):
+                        parent.getparent().remove(parent)
+                    else:
+                        parent.remove(el)
+                    found = True
+    return found
+
+
+def main() -> int:
+    ap = argparse.ArgumentParser(
+        description="List, add, or delete comments in a .docx.")
+    sub = ap.add_subparsers(dest="cmd", required=True)
+
+    p = sub.add_parser("list", help="list comments as JSON")
+    p.add_argument("path", help="input .docx")
+
+    p = sub.add_parser("add", help="add a comment anchored to text")
+    p.add_argument("path", help="input .docx")
+    p.add_argument("-o", "--output", help="output path (default: in place)")
+    p.add_argument("--target", required=True,
+                   help="anchor: first occurrence of this text")
+    p.add_argument("--text", required=True, help="comment body")
+    p.add_argument("--author", default="Qubicl")
+    p.add_argument("--initials", default="")
+    p.add_argument("--xml", action="store_true",
+                   help="force the XML fallback (skip native API)")
+
+    p = sub.add_parser("delete", help="delete a comment by id")
+    p.add_argument("path", help="input .docx")
+    p.add_argument("-o", "--output", help="output path (default: in place)")
+    p.add_argument("--id", required=True, help="comment id")
+
+    args = ap.parse_args()
+    doc = Document(args.path)
+
+    if args.cmd == "list":
+        print(json.dumps({"ok": True, "comments": list_comments(doc)},
+                         ensure_ascii=False))
+        return 0
+
+    if args.cmd == "add":
+        para, runs = find_anchor_runs(doc, args.target)
+        if not runs:
+            print(json.dumps({"ok": False,
+                              "error": f"target not found: {args.target}"}))
+            return 1
+        native = hasattr(doc, "add_comment") and not args.xml
+        if native:
+            cid = add_comment_native(doc, runs, args.text, args.author,
+                                     args.initials)
+        else:
+            cid = add_comment_xml(doc, runs, args.text, args.author,
+                                  args.initials)
+        result = {"ok": True, "comment_id": cid,
+                  "native_api": native, "anchored_to": args.target}
+    else:  # delete
+        if not delete_comment(doc, args.id):
+            print(json.dumps({"ok": False,
+                              "error": f"no comment with id {args.id}"}))
+            return 1
+        result = {"ok": True, "deleted_id": args.id}
+
+    out = args.output or args.path
+    doc.save(out)
+    result["output"] = out
+    print(json.dumps(result, ensure_ascii=False))
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

package/dist/assets/computer/skills/core/docx/scripts/docx_common.py

@@ -0,0 +1,94 @@
+#!/usr/bin/env python3
+# MIT License. Shared helpers for the docx skill scripts.
+"""Shared helpers: paragraph iteration and run-preserving text replacement."""
+from __future__ import annotations
+
+
+def iter_all_paragraphs(doc, include_headers_footers: bool = True):
+    """Yield every paragraph in body, tables (recursively), headers, footers."""
+    yield from _iter_container(doc)
+    if include_headers_footers:
+        for section in doc.sections:
+            for part in (
+                section.header, section.footer,
+                section.first_page_header, section.first_page_footer,
+                section.even_page_header, section.even_page_footer,
+            ):
+                if part is not None:
+                    yield from _iter_container(part)
+
+
+def _iter_container(container):
+    for para in container.paragraphs:
+        yield para
+    for table in container.tables:
+        yield from _iter_table(table)
+
+
+def _iter_table(table):
+    for row in table.rows:
+        for cell in row.cells:
+            for para in cell.paragraphs:
+                yield para
+            for nested in cell.tables:
+                yield from _iter_table(nested)
+
+
+def iter_part_roots(doc):
+    """Yield the XML root of the body plus every header/footer part."""
+    yield doc.element.body
+    seen = set()
+    for section in doc.sections:
+        for part in (
+            section.header, section.footer,
+            section.first_page_header, section.first_page_footer,
+            section.even_page_header, section.even_page_footer,
+        ):
+            if part is not None and id(part._element) not in seen:
+                seen.add(id(part._element))
+                yield part._element
+
+
+def replace_in_paragraph(para, old: str, new: str) -> int:
+    """Replace `old` with `new` in a paragraph, preserving run formatting.
+
+    Strategy: first replace occurrences fully contained in a single run
+    (formatting fully preserved). If the needle spans multiple runs, the
+    matched runs are collapsed: the replacement inherits the formatting of
+    the run where the match starts. Returns number of replacements made.
+    """
+    if not old or old not in para.text:
+        return 0
+    count = 0
+    # Pass 1: within-run replacements.
+    for run in para.runs:
+        if old in run.text:
+            count += run.text.count(old)
+            run.text = run.text.replace(old, new)
+    # Pass 2: cross-run occurrences.
+    while old in para.text:
+        runs = para.runs
+        # Map paragraph text offsets to (run_index, offset_in_run).
+        full = "".join(r.text for r in runs)
+        start = full.find(old)
+        if start < 0:
+            break
+        end = start + len(old)
+        pos = 0
+        spans = []  # (run_idx, cut_start, cut_end) portions inside the match
+        for i, r in enumerate(runs):
+            r_start, r_end = pos, pos + len(r.text)
+            if r_end > start and r_start < end:
+                spans.append((i, max(start, r_start) - r_start,
+                              min(end, r_end) - r_start))
+            pos = r_end
+        first = True
+        for i, cs, ce in spans:
+            t = runs[i].text
+            if first:
+                runs[i].text = t[:cs] + new + t[ce:]
+                first = False
+            else:
+                runs[i].text = t[:cs] + t[ce:]
+        count += 1
+    return count