Command Claude Code
Verify version.json, .claude/framework-manifest.json, .claude/framework-installed.json, and install.ps1 header agree on the installed version (trust order = version.json wins).
Command Claude Code
The meta-lockstep enforcer — couples a scan's cross-provider SCOPE FILTER to the classderivation rule-table, so broadening one without the paired rule is self-detecting. Report-only in /scan:full.
Command Claude Code
The named enforcer for the role-registry model/effort CHAIN. Report-only in /scan:full.
Command Claude Code
Read-only diagnostic — list Node processes on the host with PID, start-time, working-set KB, and command.
Command Claude Code
Wiring gate for the operator message contract — proves the Stop hook (operator-contract-lint.js) is registered, compiled into settings.json, its fixture suite is green, and the three doctrine homes (paths.operatorContract, CLAUDE.md, the lib's HEADERS constant) have not drifted. Fail-closed; report-only 7-day…
Command Claude Code
Coverage enforcer for the panel-registry (the /panel: suite) — every panels row is well-shaped ({name, opener, description, runcontext}; runcontext ∈ {inapp, cli}) and its opener resolves to a real backing target (a node file exists; a /ns:name skill resolves via dispatch-skill --resolve). An orphan/phantom/unsafe…
Command Claude Code
Cross-run intelligence and automation proposals — diagnose recurring patterns or propose prevention.
Command Claude Code
Report-only plan-lint — flags any plan artifact under planning/epics/ (optionally planning/plans/) that omits a principle-required section: a named enforcer, proof/acceptance, or a blast-radius assessment. The named enforcer for the planning principles (the planning principles doc). Fail-open; never blocks.
Command Claude Code
Pre-publish scan for personal data — credentials, emails, homedir paths, runtime files tracked by git.
Command Claude Code
The carve-out enforcer — a provider != claude role must NOT carry Agent-tool reachability (tools:["Agent"] in the registry or its spec frontmatter), because the harness Agent tool is Claude-only. Report-only in /scan:full.
Command Claude Code
Cross-file reference integrity — broken links, orphans, stale SPECGRAPH edges.
Command Claude Code
Run the regression-seed suite — the recurring bug classes, made runnable. Reports per-class pass/fail/gap + catch-rate. Framework repository only (an install does not carry the suite).
Command Claude Code
Specification consistency, coverage, and drift — static audit, change-driven propagation check, or pending-drift review.
Command Claude Code
Assert every done/retrospected/released sprint has BOTH a Sprints-table ledger row AND a Shipped narrative entry in ROADMAP.md — covers the narrative half of /sprint:full Step 8b.
Command Claude Code
The one check that owns role parity across the org map, the dispatch catalog, and team-guard — fail-closed enforcer so repartitioning agents by domain can't silently drift the registry.
Command Claude Code
Verify the MC app scaffold (Next+Tailwind v4+shadcn/ui+Radix+Lucide) is complete and coherent — fail-closed enforcer, so every scaffolded product ships a real component library, not vibe-coded raw elements.
Command Claude Code
Scan-suite self-inventory — asserts every /scan: skill is delegated by /scan:full or explicitly excluded (with a reason). Kills the dir↔aggregator drift class.
Command Claude Code
The security-binding-lane enforcer — the security BINDING verdict must resolve to a verifiable provider, never agy, while served-model proof is open, plus catalog↔providers alias-inclusive DEFAULTPROVIDER consistency. Blocking in /scan:full.
Command Claude Code
Bidirectional coverage of the skill hook-point registry — REVERSE (registry coherent vs role-registry) + FORWARD (every registered skill has a command file) + HARDCODE/STALE (two asymmetric checks — no skill body names a renamed-away management persona ANYWHERE, and no skill dispatches a current persona via…
Command Claude Code
Audits Beta consultation honesty across /sprint:full runs (missing consults, placeholder verdicts, ESCALATE-without-halt).
Command Claude Code
Bidirectional coverage of the sprint hook-point registry — FORWARD (every matched block-row has a managerconsult record per /sprint:full run) + REVERSE (registry structurally coherent against role-registry).
Command Claude Code
Audits manager-consult coverage across /sprint:full runs — asserts the design-quality authority was consulted on every UI/app-design/web-design-touching sprint.
Command Claude Code
System inventory — enumerate every active MC system, diff against manifest, report drift and gaps.
Command Claude Code
Reconstruct a build timeline from transaction, event, and provider logs.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: