This release includes breaking changes for platform teams planning a safe upgrade.
✓ No known CVEs patched in this version
Topics
Summary
AI summaryFixed lint false positives for fallback trailers and forward-reference diagnostics, reducing noise.
Full changelog
Two mechanical lint fixes from Perry's R-series next-ring (77ed6c65).
Standalone patch ahead of the AgentConnector audit work — keeps
contract-design surface separate from cold-author polish per the
thematic-discipline pattern that worked through v0.9.x. No contract
surface touched; v1.0 cold-author signoff (mean UX 4.0/5 against
v0.9.4) still stands as the gate-1 baseline.
Fixed — fallback-trailer lint-layering
Cold authors expected (fallback: ...) on a bare $ TOOL op to
suppress the tier-1 unwired-primary-connector error (the intuition:
"if I provide a fallback, the runtime is guarded"). It didn't — lint
runs at authoring-time and fallback resolves at dispatch-time, so the
error fired regardless of fallback presence. The layering wasn't
obvious from the diagnostic.
Fix: when every call site for a (target, tool) pair carries an
op-level (fallback: ...) trailer, demote the finding from error
to info. The advisory message explains the layering ("Lint runs at
authoring-time and the fallback resolves at dispatch-time; the
runtime branch is reached only if dispatch fails in production"). If
any call site lacks a fallback, the error stays. Per R8 cold-author
finding in 77ed6c65.
Fixed — forward-reference noise dedup
Pre-v0.9.5 lint emitted up to 4 diagnostics for 2 missing skills:
unknown-skill-reference keyed by via:name (so the same missing
skill referenced via both & and $ execute_skill produced 2
findings) AND the paired deferred-skill-reference advisory doubled
the count. Per Perry's "4 diagnostics for 2 missing skills" cold-author
finding — noisy without adding signal.
Fix:
unknown-skill-referencenow dedups by skill name (notvia:name);
the message lists all vias encountered.deferred-skill-referenceadvisory removed entirely — the warning's
remediation field already explains the forward-ref path. One warning
per missing skill carries the same signal.
Net: 2 missing skills × dual-ref produces 2 findings (down from 8).
Cold author noise floor reduced; pre-adoption rule applies (no
external migration needed).
Weekly OSS security release digest.
The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.
No spam, unsubscribe anytime.
Share this release
About Skillscript
All releases →Related context
Related tools
Earlier breaking changes
- v0.35.0 Editing `# Deadline:` on an approved skill drops it to Draft (signature invalidates).
Beta — feedback welcome: [email protected]