Skip to content
user

converters

Document conversion — between every format your team uses.

8 skills1 human gate5–15 min/session
Install this pack Read the reference

What changes when you install this

After installing converters, document conversion is a named operation with a verifiable output. `file-to-markdown` converts PDFs, DOCX, and images to clean Markdown. `msg-to-markdown` converts Outlook .msg and MIME .eml email. `mermaid-renderer` bakes Mermaid fence blocks to SVG or PNG. The four Markdown-to-* skills fill a named template and produce styled office output. `render-proof` renders a Markdown draft as an offline HTML proof for human review. A three-tier capability model ensures the floor works in locked-down environments.

Skills in this pack

  • file-to-markdown

    Converts documents — PDF, DOCX, PPTX, XLSX, and images — to clean Markdown using a three-tier capability model (no-dep, local-tools, agent-vision).

  • msg-to-markdown

    Converts Outlook .msg and MIME .eml email to Markdown, preserving headers, body structure, and attachment inventory.

  • render-proof

    Renders a Markdown file as a self-contained offline HTML proof artifact for human review — muted palette deliberately distinct from the publication look.

  • markdown-to-html

    Produces a styled, self-contained HTML page from a Markdown source — shareable without external assets.

  • markdown-to-docx

    Fills a branded Word template from a Markdown source — brand intact, no manual reformatting.

  • markdown-to-pptx

    Fills a branded PowerPoint template from a Markdown source — slide structure derived from headings.

  • markdown-to-xlsx

    Fills a branded Excel template from a Markdown source — tables and structured data mapped to sheets.

  • mermaid-renderer

    Renders Mermaid fence blocks from a Markdown file to SVG or PNG — for tools that don't render Mermaid live.

The journey

Stage 1 — Invoke the converter

You name the source file and the target format. The agent invokes the appropriate skill: file-to-markdown for a PDF or Office document, markdown-to-docx or markdown-to-pptx for output artifacts, mermaid-renderer for diagrams embedded in a Markdown file.

You: Name the source file explicitly and confirm the target format. For office output, also name which template to use — the agent cannot discover your organization’s brand template without you pointing to it. If the source is a large PDF and you only need a specific section, say so before the agent converts the whole document.

Stage 2 — Tier selection (for file-to-markdown)

For document extraction, the agent selects the appropriate capability tier: Tier 1 (no external tools — pure text extraction), Tier 2 (local tools like Pandoc or Poppler), or Tier 3 (agent vision for scanned or image-heavy documents). It reports which tier was used and why.

You: Confirm the tier makes sense for the document type. If the document is a scanned PDF and the agent selects Tier 1, the extracted Markdown will contain garbled text — redirect to Tier 3 or flag the document as needing OCR before extraction.

Stage 3 — Review the output

The agent reports where the output lands and emits a brief summary of what it converted.

You: Open the output file and review it at the G-output gate. The agent’s report says “conversion complete” — what it cannot tell you is whether the resulting file is what you needed. Check for truncation, structure loss, and template application. If a section is missing, re-run with an explicit instruction to include it.

Human gates

For each gate, everything you need to make a confident decision.

  • G-output

    Review the converted output

    Trigger
    After any converter skill reports completion
    Time
    2–5 minutes
    What to check, good, bad, consequence

    What to check

    • Is all key content present — no silent truncation of large sections or long tables?
    • For office output: does the template apply correctly — brand, fonts, and structure intact, not plain Times New Roman?
    • For Markdown extraction: are code blocks, tables, and document structure preserved as intended?
    • For mermaid-renderer: does the rendered image match the diagram in the source — no missing nodes or broken edges?

    What good looks like

    An output that faithfully represents the source content in the target format, with structure and styling preserved — something you'd be comfortable sending to a stakeholder.

    What bad looks like

    An output that passes the file-exists check but silently dropped a section or table. Or office output where the template wasn't applied — the agent reported success but the file opened in the wrong font.

    Consequence of skipping

    Converter outputs are usually shared with stakeholders or used as input to another workflow. A silently incomplete conversion is discovered by the stakeholder, not the agent — and by then it's too late to re-run quietly.

Typical session

Agent turns
2–4
Human gates
1
Wall-clock time
5–15 min

Install and next steps

agentbundle install --pack converters --scope user