SOP Generator Skill
Goal
Convert an unstructured transcript into a professional SOP.
Workflow
- Read the transcript file provided by the user.
- Clean it:
python scripts/preprocess.py <transcript-file>. - Extract objective, prerequisites, ordered actions, and expected result, following references/extraction-rules.md.
- Fill in references/sop-template.md, following references/writing-guidelines.md.
- Output the finished SOP as Markdown in the chat response.
- If the user asks for a DOCX, save the Markdown SOP to a file (e.g.
SOP.md) and run:python scripts/export_docx.py SOP.md Generated_SOP.docx.
Rules
- Never include speaker names.
- Never include timestamps.
- Never invent information not present in the transcript.
- Use numbered steps in the Procedure section.
- Use Markdown output, matching the section order in references/sop-template.md exactly, so output is consistent across runs and machines.
Test
Run this skill against each pair below and compare the result to the matching expected output. Structure and step order should match; exact wording may vary slightly between runs.
- assets/sample-transcript.txt → assets/expected-output.md (linear deployment SOP)
- assets/sample-transcript-support.txt → assets/expected-output-support.md (two-party troubleshooting SOP with a conditional escalation path)