pmf-import

name	pmf-import
description	One-time import of an existing HTML/JSON PMF tracker (a pmf-data.json export) into the Markdown tracker layout — preserving the overview, thesis, scratchpad, every hypothesis, and weekly reviews verbatim. Use when migrating from a prior tracker, or when the user says 'pmf import', 'import my pmf data', 'migrate the tracker', 'convert pmf-data.json', or types '/pmf-import'.
argument-hint	[--data <pmf-data.json>] [--out <pmf/ root>]
version	0.1.0
user_invocable	true

/pmf-import — migrate an existing tracker into Markdown

This wraps the deterministic importer (scripts/pmf_import.py) with a safe, confirmed flow. The script does the lossless field mapping; you handle path resolution, the safety confirmation, and the summary.

Before you do anything

1. Read ${CLAUDE_PLUGIN_ROOT}/references/data-model.md (the field-by-field contract) and honesty-rules.md.
2. Resolve inputs:
   • --data = the source pmf-data.json. If the user didn't give a path, ask (AskUserQuestion) for it.
   • --out = the target tracker root $PMF (default pmf/ under git rev-parse --show-toplevel).
   • Always confirm the source path with the user — never assume a default location.

Steps

1. Dry run first — show the plan, write nothing:

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/pmf_import.py --data "<data>" --out "$PMF" --dry-run
   

2. Confirm via AskUserQuestion: "Create N new files (skip M existing)?" — show the create/skip counts from the dry run.
3. Import (safe — never overwrites):

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/pmf_import.py --data "<data>" --out "$PMF"
   

   Existing files are skipped and reported; use --force only against a throwaway directory.
4. Verify the round-trip:

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/pmf_import.py --data "<data>" --out "$PMF" --verify
   

   This asserts every hypothesis and review field matches the source verbatim (empty result → TBD). Surface any mismatch.
5. Build the index:

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/pmf_index.py --out "$PMF" --check
   

6. Report what was created, the verify result, and remind the user that imported hypotheses carry an evidence_weight of 🔵 and empty sources (auto-inferred on import — there's an HTML comment in each file flagging this). They should review and correct those during the next /pmf-wrap.

Honesty note

The importer adds an explicit "auto-inferred on import — verify" marker for any field that has no source in the original data (evidence_weight, sources, created/updated). Never silently treat those as real provenance. Everything else is carried over verbatim.