teststand-mcp: Instructions file for Claude Code

TestStandMCP — Behavior Rules for Claude

Rebuilding a .seq 1:1 — export_sequence_file + import_sequence_file FIRST

For a whole-file reproduction, migration or bulk edit, use the export/import pair. It is the default path; the granular tools are for surgical single edits.

export_sequence_file(file_path)                    → writes <file>.model.json, returns a summary
create_sequence_file(dest, overwrite=true)         → overwrite handles the close+delete dance
import_sequence_file(model_path, dest_file_path)   → rebuilds everything, returns counts + warnings[]
diff_sequence_files(orig, dest)                    → verify the CONTENT (rows capped at 150 by default)
audit_type_consistency(dest, reference_file_path=orig) → verify the TYPE REGISTRY (the diff can't)

The import removes the destination's leftover MainSequence itself when the model has none, so there is no delete_sequence step any more.

Measured 2026-07-29 on 8 sequences of TFW_MDC_com_Python.seq (47 steps, 13 object-oriented Python steps, 5 LabVIEW .lvlibp steps, 1 cross-file SequenceCall): 3 MCP calls and ZERO FileDiffer differences inside the imported scope (the only rows left are the 22 sequences not imported). The same rebuild with the granular tools took ~700 calls, 3 diff iterations and left 224 differences.

The on-disk FORMAT is reproduced too — and the diff is blind to it (2026-07-31)

A .seq is stored either as compressed binary (TOF1 magic, zlib body — step names are NOT text-searchable) or as XML (a UTF-8 BOM then <?xml), and the engine's default for a new file is binary. MEDELA_TFW's real files are XML, so a rebuild used to come out binary: content-identical (diff_sequence_files says identical) yet different in every byte, 25 KB against 3.4 MB on TFW_MDC_com_Python — ×133, pure serialization.

• The export now captures the source format (file.fileFormat in the model) and the import reproduces it by default — nothing to pass for a 1:1 rebuild; the outcome reports fileFormat.
• get_file_properties reports fileFormat; create_sequence_file, save_sequence_file, set_file_properties and import_sequence_file take file_format = binary|xml|ini. The format is stored IN the file, so it survives; an ordinary save never converts it back.
• Before reading a large size drop as data loss, check the first bytes (TOF1 vs <?xml).
• XML is the format to pick for anything that lives in git — binary .seq diffs are opaque.
• This is the serialization, NOT the TestStand version target (that is the version field).