v1.0.0-beta.5

Highlights

New commands: compile deterministic terminology edits, check-spec-freshness stale spec detection.

Semantic runtime expansion: check-compliance, check-terminology, and canonicalize now support LLM-backed adjudication via runtime.analysis.

MCP surface doubled (6 → 13 tools): check_compliance, check_terminology, governed_by, compile_preview, fix_preview, status, explain_file. Agents get the full code-development loop: look up governing specs → write code → check compliance → verify terminology → preview fixes.

Parallel LLM adjudication: check-doc-drift and check-compliance now run LLM calls concurrently (bounded errgroup, concurrency 4). Wall-clock time scales inversely with concurrency.

UX improvements: ranked analyze-impact --summary mode, overlap pre-check, impact severity classification (breaking/behavioral/cosmetic), JSON rebuild progress events, multirepo explain-file and status path improvements.

What's Changed

• docs: add semantic runtime operator note by @autholykos in #214
• UX strengthening slice (ranked summaries, json progress, multirepo paths) by @autholykos in #226
• feat(skill): autoskiller v8 pituitary-cli optimization by @autholykos in #231
• feat: extend semantic runtime to check-compliance, check-terminology, and canonicalize by @autholykos in #233
• feat: add new overlap pre-check and analyze-impact severity classification by @autholykos in #233
• feat: add pituitary compile command by @autholykos in #234
• feat: add check-spec-freshness command by @autholykos in #240
• feat: expand MCP server tools for code-development workflow by @autholykos in #241
• feat: parallelize LLM adjudication in check-doc-drift and check-compliance by @autholykos in #242

Full Changelog: v1.0.0-beta.4...v1.0.0-beta.5