[bcastets-robotiq]

Problem

The daily update-readme run failed (run):

Could not import the "ROS2" software tools section: no AUTO-GENERATED-ROS2-TABLE markers found in robotiq.github.io's docs/intro.mdx. It may have been restructured — refusing to publish a partial section.

Correct behavior — the safety check from PR #2 did exactly its job and refused to publish a broken section. The docs site restructured again:

• ROS2 and ROS1 merged into a single ROS category (one column, not per-distro).
• Each section's own ### Heading moved from outside its marker pair to inside it (previously: heading, blank line, {/* START */}; now: {/* START */}, heading, blank line, table).

The second part is a real bug fix, not just a config update: naively updating SOFTWARE_SECTIONS without handling the heading move would have produced a duplicated #### ROS / ### ROS pair in the README (confirmed by actually running the regenerator before this fix).

Fix

• SOFTWARE_SECTIONS: drop ROS2/ROS1, add single ROS.
• extractMarkerBlock now strips a leading heading line from the extracted block regardless of which side of the marker it's on, since buildSoftwareToolsSection always supplies its own heading.
• Refreshed scripts/fixtures/intro.mdx from the live site; updated test expectations to match (new heading list, new product names, new Libraries/ROS paths).
• Added a regression test for the inline-heading stripping.

Review round 2 (ebarnett3, mbegin-robotiq)

• Fixed: the heading-strip regex required a trailing newline after the heading, so a block that's nothing but a heading (no other content) passed through unstripped, still leaking a duplicate heading. Fixed per the suggested /^#{1,6}[^\n]*(\n+|$)/, with a regression test for exactly that case.
• Fixed: strengthened the fixture test with an end-to-end assertion (assert.doesNotMatch(section, /^#{1,3} /m)) that no stray upstream heading leaks through — the previous check only looked for our own #### headings and would've passed even if the strip logic regressed.
• Deferred, not fixed here: 7 grammar nits in the fixture — all copied verbatim from the live robotiq.github.io page, not something this repo controls. Filed as robotiq/robotiq.github.io#36; replied individually on each thread.

Test plan

• All 18 tests pass (node --test scripts/*.test.mjs)
• Ran the script end-to-end against live data — output verified clean (no duplicate headings, correct single ROS table)

🤖 Generated with Claude Code

[mbegin-robotiq]

LGTM. One non-blocking suggestion inline about test coverage.

[bcastets-robotiq]

Pushed fixes for both actionable review comments (heading-only edge case + the stronger fixture assertion) in 2253eaa.

Filed robotiq/robotiq.github.io#36 for the grammar nits — those are in the live docs page's own text (copied verbatim into the fixture), not something to fix here.