Multi-Artifact Work Package
Takes a business situation or client brief and produces a defined set of real deliverables with an artifact contract that specifies exactly what each file must contain. The core problem it solves: AI models frequently produce the right number of files with the right names but wrong formats — a PowerPoint that is actually HTML, a spreadsheet with no formulas, a PDF that is just a renamed text file. This skill makes the artifact contract explicit before generation begins and verifies compliance after.
Trigger
Use when the user says "/multi-artifact-work-package", "build a deliverable package", "create a full set of client outputs", "generate artifacts for this brief", or when any task requires more than two coordinated output files in different formats.
Phase 1: Situation Intake
Ask the user for:
- Business situation or brief — free-form description of the problem, client, and goals.
- Audience — who receives the deliverables (executives, board, client, internal team).
- Output directory — where to write the files.
If the user provides a document or notes, read them before asking clarifying questions.
Phase 2: Artifact Contract Definition
Before any generation, define the artifact contract explicitly. Produce a contract in this format and ask the user to confirm or edit it before proceeding:
ARTIFACT CONTRACT
-----------------
1. executive-summary.pptx
Format: Real PowerPoint (python-pptx)
Required: ≥ 5 slides, embedded chart, title + body text per slide
Audience: C-suite
2. financial-model.xlsx
Format: Real Excel (openpyxl)
Required: 3 sheets, ≥ 10 formula cells, named ranges, chart on Sheet 3
Audience: CFO / analyst
3. full-report.pdf
Format: PDF with selectable text (reportlab or weasyprint)
Required: ≥ 5 pages, table of contents, section headers
Audience: Full team
4. press-release.docx
Format: Real Word document (python-docx)
Required: AP-style, dateline, quote block, boilerplate footer
Audience: Media
5. faq.md
Format: Markdown
Required: ≥ 8 Q&A pairs
Audience: Internal
Do not begin generation until the user approves the contract. This is the HIL gate.
Phase 3: Generation
For each artifact in the approved contract, generate using the appropriate library:
- PowerPoint: Use
python-pptx. Never write HTML and rename it. Every slide must have ashapes.titleandtext_framewith real content. - Excel: Use
openpyxl. Every "formula" cell must contain an=expression, not a hard-coded number. Include at least one chart object. - PDF: Use
reportlaborweasyprint. Output must start with%PDF-bytes. Never rename an HTML file to .pdf. - Word: Use
python-docx. Output ZIP structure must containword/document.xml. - Markdown / plain text: Standard file write.
After writing each file, print its path and size.
Phase 4: Verification
After all files are written, run artifact-contract-validator checks inline:
- Open each file and verify format authenticity (magic bytes / ZIP structure).
- Verify required contents (formula count, slide count, page count).
- Produce a pass/fail line per file.
If any file fails, regenerate that specific artifact and re-verify. Do not proceed to delivery until all contract items pass.
Phase 5: Delivery Summary
DELIVERY SUMMARY
================
Directory: {path}
Generated: {date}
Artifacts
---------
{filename} {size} PASS
{filename} {size} PASS
All {n} artifacts passed contract validation. Ready for delivery.
What This Does NOT Do
- Does not validate content accuracy, legal correctness, or data sourcing.
- Does not handle deliverables requiring live data connections.
- Does not replace
/artifact-contract-validatorfor post-hoc audits of externally-generated files.
Dependencies
/artifact-contract-validator— called inline in Phase 4; install first if running standalone.
Source
Extracted from Nate Kadlac newsletter (2026-04-28): "ChatGPT 5.5 scored 87 where the next best model scored 67. Here's what that gap looks like in real work." Prompt 2 of the five-prompt kit: Multi-Artifact Work Package, including the artifact-contract verification layer.