Resolve two consistency nits from review: shape-spec said plan-review was
'optional' for Small while the README canonical table says skip it (align
to skip); and the changelog now describes both misuse directions the new
rationalizations guard against, not just gate-skipping.
Add one rationalization row to each design-phase skill against using the
tier model as an excuse to skip the evidence gate (verification-gate /
code-review-loop never scale by tier), and record the change in the
changelog.
Add a Trivial/Small/Standard tier model so the design ceremony scales to
the change while the evidence gate never does. README gains a canonical
'Sizing the work' table + the non-negotiable rule (verification-gate and
code-review-loop always run); shape-spec gets a Step 0 sizing gate;
write-plan and plan-review state which tier they belong to.
- normalize CRLF and match the closing frontmatter delimiter tolerantly
(no false CI failures on Windows-committed files; no latching onto a
later horizontal rule)
- strip fenced code blocks before section/table scanning so embedded
anatomy examples can't fool the checks
- strip quotes/inline comments from frontmatter values; reject empty
required values
- guard readdir/stat so a wrong root or dangling symlink reports instead
of crashing; fail on zero skills found
- require plugin.json version to exist (undefined===undefined no longer
passes as in sync)
Checks every skills/*/SKILL.md for required frontmatter (name matching
directory, description, user-invocable) and the 8-section anatomy from
README (init is anatomy-exempt as the documented off-spine skill), plus
plugin.json/marketplace.json version sync. Zero deps, exits 1 on any
violation.
duthaho
