v0.6.1

Two defects in v0.6.0, both of which could cost you work

Update if you are on v0.6.0. Neither of these is theoretical — one was reproduced killing a live KiCAD, the other by a user who ran what looked like a documentation command.

KiCAD could be terminated while placing an ordinary footprint (#237)

Placing a footprint whose library drawing includes a Dwgs.User item — silkscreen-adjacent artwork that is entirely normal in the official libraries — sent KiCAD a layer value it does not validate. KiCAD 10.0.5 indexed its layer bitset with it and faulted at 0xc0000005 in kicommon.dll. The KiCAD process exited, taking any unsaved board with it. Konnect saw only an NNG receive timeout.

Two official footprints reproduce it exactly:

• Connector_USB:USB_C_Receptacle_GCT_USB4105-xx-A_16P_TopMnt_Horizontal
• Connector:BJB_Pico_46.110.1001_Receptacle_Horizontal

The USB-C part also reproduced through update_pcb_from_schematic, so any sync touching it was affected.

The cause was a layer table that stopped at In2.Cu and had no entry for Dwgs.User, Cmts.User, Eco1/2.User, F/B.Adhes, Margin, Rescue, or any of User.1–User.45. Anything unlisted became BL_UNDEFINED.

Fixed in two parts, because widening the table alone would leave the next layer KiCAD adds able to do the same thing:

• the table now covers every layer a KiCAD 10 footprint can legally draw on, with inner-copper and user layers computed rather than listed;
• an unrepresentable layer is now refused before anything is sent, naming the layer and the item. build_footprint_item validates the root layer, every pad layer and every graphic layer up front.

Reported with a complete root-cause analysis by @neusse, including both fixtures and the faulting address. The widened mapping follows the arithmetic in @TWT233's #232, which reached the same table independently.

konnect init --help ran the installer (#238)

konnect init --help did not print help. It installed, overwriting whatever you had under ~/.claude/skills, ~/.claude/agents and ~/.claude/settings.json. konnect uninstall --help reached the same code and would have removed them.

The subcommands were matched before the --help arm, and the client parser silently skipped every argument that was not --client.

--help and -h are now answered wherever they appear, before dispatch, with help for the subcommand they accompany. Verified against a real home directory: 26 watched files, zero changed.

Also stricter, and worth knowing about: an argument Konnect does not recognise is now an error instead of being ignored. konnect init --cleint codex used to install for Claude — the default — which is a surprising thing to find halfway through setting up Codex. It now exits non-zero.

If you launch Konnect as an MCP server with a flag other than --config or --client, it will now refuse to start rather than ignoring it. Those two are the only options the server has ever read, and the bundled examples/*.json use only --config.

Known limitations

• After KiCAD crashes for any reason, the next closed-board call still edits the file. Once the transport is unreachable, place_component's fallback cannot tell "KiCAD was never running" from "KiCAD died a second ago holding this board". Found while verifying #237 and filed as #240; unchanged in this release.
• moving_and_rotating_footprint_preserves_child_geometry remains flaky in the manual live suite (#221).
• Closed-board move_component / rotate_component (#228) and flip_component (#115) are in review, not in this release.

Unchanged

No tool was added, removed or renamed, and no tool's response changed shape. Still 202 tools across 19 toolsets, 208 with meta-tools. Board files and schematics are untouched by both fixes.

---

Full Changelog: v0.6.0...v0.6.1