메뉴
Build or edit slide decks (.pptx) as real files — pitch decks, briefings, status readouts — or extract content from existing ones. Use whenever slides or presentations are the input or the deliverable. Not for documents, spreadsheets, or PDFs.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
| name | pptx |
| description | Build or edit slide decks (.pptx) as real files — pitch decks, briefings, status readouts — or extract content from existing ones. Use whenever slides or presentations are the input or the deliverable. Not for documents, spreadsheets, or PDFs. |
Write the deck with execute_code using python-pptx, save into the fenced
project folder (or configured output_dir), report the path.
Draft the deck as a text outline (title + bullets per slide), confirm it with the operator if scope is at all unclear, then render. Rendering is the cheap part; the outline is where the deck is won.
from pptx import Presentation
from pptx.util import Inches, Pt
prs = Presentation() # or Presentation(template_path) to inherit branding
title_slide = prs.slides.add_slide(prs.slide_layouts[0])
title_slide.shapes.title.text = "Q3 Ops Review"
title_slide.placeholders[1].text = "Prepared by …"
for heading, bullets in outline:
slide = prs.slides.add_slide(prs.slide_layouts[1])
slide.shapes.title.text = heading
body = slide.placeholders[1].text_frame
body.text = bullets[0]
for b in bullets[1:]:
body.add_paragraph().text = b
prs.save(path)
slide.notes_slide.notes_text_frame.text = "…" — offer
them; presenters love not writing them.slide.shapes.add_picture(...); for data charts, render
with a plotting library to PNG first, then place.Iterate prs.slides → slide.shapes → shape.text_frame for extraction
(e.g. turning a deck into a summary). Edits: modify in place, save to a new
file unless in-place was requested.
Re-open and count slides; read back the titles. Report "12 slides, saved to " — verified, not assumed.
Once the file is written and verified, register it in the Artifact panel with
save_file_artifact(path, title=…) when that tool is available (the artifact plugin, protoAgent
v0.107.0+). It stores the bytes, shows a Download card with a readable preview (a slide outline), and keeps
an edit history — pass the same artifact_id to save a later revision as v2/v3. Report the saved
path and that it's in the panel. If the tool isn't available, just report the path.
Produce or modify Word documents (.docx) as real files — reports, memos, letters, proposals, or any deliverable the operator will open in Word. Use when the requested output is a Word file, or when an existing .docx needs content extracted, edited, or restructured. Not for spreadsheets, decks, or PDFs — those have their own skills.
Read, extract from, combine, split, watermark, fill, or produce PDF files. Use whenever a .pdf is the input or the deliverable — merging reports, pulling text or tables out, rotating scans, filling forms, or generating a polished PDF from content.
Produce, edit, or clean spreadsheets (.xlsx/.csv/.tsv) as real files — reconciliations, trackers, budgets, data cleanups, format conversions. Use whenever the deliverable is a spreadsheet or an existing one needs work, including messy tabular data that needs restructuring. Not for Word/PDF/deck outputs.
Compose the operator's daily brief — schedule, things waiting on them, ongoing work, and heads-ups — rendered as a polished HTML artifact, optionally on a recurring schedule. Run only when the operator explicitly asks for their brief (or /daily-brief), or when executing a scheduled brief task. An ordinary question about today's calendar is answered directly, not with a brief.
Conversational first-run companion for the Cowork archetype — verify the workspace is ready, import existing Claude Code/Cowork state, prove the setup on a real task, learn the operator's writing voice, and wire the first scheduled task. The mechanical setup (archetype, plugins, dependencies, folders) belongs to the setup wizard; this skill picks up where it leaves off.
Reflective maintenance pass over long-term memory — merge duplicates, retire stale or completed items, sharpen what stays. Use when the operator asks to clean up/organize memory, when recall keeps surfacing outdated facts, or periodically as a scheduled task.