KiCAD-Prism (krishna-swaroop) errors

16 recurring problems engineers report in KiCAD-Prism (krishna-swaroop), ranked by impact.

#21 by errors trackedECADPLMMCAD

Recurring KiCAD-Prism (krishna-swaroop) errors, by impact

  1. 1A fitted part with neither an MPN nor an LCSC number was never stored as a BOM line. On a USB-PD trigger board, connector J2 disappeared, and the supply page showed '17 of 17 lines', 100% coverage and '10 of 10 boards' buildable. The fix counts it as unchecked, so coverage now reads 94%.ECAD · Agent failure · 1 report74
  2. 2A release BOM over 20,000 rows was silently cut to a prefix and assessed as if complete; DeepSeek did the first pass of the fixes and the author corrected two of its changes.ECAD · Agent failure · 1 report74
  3. 3Alternate parts were approved without recording which main part they were reviewed against, so a new reviewed_against column now shows 'vs NEW, was vs OLD' when the main part changed.PLM · Agent failure · 1 report74
  4. 4For repos with several KiCad projects in subfolders, every jobset job succeeds but nothing gets committed; the log ends with a 'Git sync warning', outputs sit untracked, and the Assets Portal shows 'No files found' for design and manufacturing outputs. kicad-cli also rewrote the .kicad_prl during the run, so the checkout is dirty and the next Sync cannot fast-forward.ECAD · Agent failure · 1 report74
  5. 5is_fitted only matched 'DNP = yes' while KiCad's BOM export writes DNP, Excluded from BOM, 1 or x across several columns, and BOM lines with no tracked part or stock reading were counted as fine.ECAD · Agent failure · 1 report74
  6. 6When the board is behind its schematic, two distinct schematic nets with pins on the same stale board net got folded together and inspected under the board's old name. The fix only applies when more than one schematic net claims a board net.ECAD · Agent failure · 1 report74
  7. 7A failed supply read used to look like empty data ('No rates stored yet'), unknown stock read inconsistently, and 'Nothing to review' used a different count from the badge.PLM · Agent failure · 1 report66
  8. 8a production rebuild of a board's design bundle failed with an unbalanced parenthesis at byte 49,020,829 inside a 15.8 MB ground zone; both retries resumed from the same bad archive and failed the same way, while a fresh rebuild of the same commit minutes later passed, cause never found.MCAD · Agent failure · 1 report66
  9. 9a real project named its child sheets "subsheets/" while the folder was "Subsheets/"; the core design found them but project refs and the semantic index each had their own path logic, so project.json listed only the root sheet and the viewer showed one page instead of all 22 subsheets.ECAD · Agent failure · 1 report66
  10. 10After deploying v5.0.0-alpha, real KiCad designs got no schematic model: KiCad 9 private fields failed the whole schematic, a case-differing sub-sheet name failed on the Linux worker, and two mounting pads sharing a pin number got the same pin UUID, breaking the compiled graph.ECAD · Agent failure · 1 report66
  11. 11an independent KiCad 10 parser compared against kicad-cli diverged on duplicate symbol UUIDs collapsing occurrences, bus joins and same-name labels going wrong, and multi-root projects not being read; after fixes 202 of 206 corpus projects match, four still do not, and on some KiCad itself varies run to run.ECAD · Agent failure · 1 report66
  12. 12Each supply doc page was checked claim by claim against the branch and corrected, including a wrong status rule and a wrong 'paused for 12 hours' claim.PLM · Agent failure · 1 report66
  13. 13Panels get made with KiKit outside the release tool, inspected in KiCad, and their settings and files managed separately from the revision and variant that produced them. The ask is to tie the panel to the release.ECAD · Agent failure · 1 report66
  14. 14The KiCad panel let QA read a part's supply status, but the REST route and MCP tool required designer or admin rights, so the same action had different permissions by door.PLM · Agent failure · 1 report66
  15. 15The parts catalog stored a symbol extending another without its parent, so KiCad refused to load it, and the release route checked read access only before moving a component.ECAD · Agent failure · 1 report66
  16. 16two design-bundle failures retried forever: a copy-paste left two units of one symbol sharing a UUID so verification refused the archive, and an LFS-tracked STEP model on a branch untouched for a week had no local object because sync only fetched LFS for refs updated in the last 7 days.ECAD · Agent failure · 1 report66

How it fails

  1. Agent failure16

Sourced from public community reports. Costs and delays on individual entries are modeled estimates, not measurements.