owlrunkit 0.18.3 → 0.19.0

package/integrations/codex/skills/owlcoda-runkit/SKILL.md

@@ -1,16 +1,16 @@
 ---
 name: owlcoda-runkit
-description: Run receipt-backed, project-owned work with OwlCoda RunKit. Use when Codex needs to coordinate a large multi-Agent project, show per-Agent progress, manage dependencies or decisions, atomically return failed work for evidence-linked rework, record intentionally deferred verification without repeating covered tests, hand work to another Agent, recover a blank-session next action, initialize or inspect `.owlcoda/runkit/`, plan bounded work with engine pins and leases, capture exact verification evidence, close an execution, produce a ready-for-commit receipt, or perform a read-only foreign-project shadow without widening Git or release authority.
+description: Run receipt-backed, project-owned work with OwlCoda RunKit. Use when Codex needs to coordinate a large multi-Agent project, show per-Agent progress, manage dependencies or decisions, atomically return failed work for evidence-linked rework, import an external DeliveryPacket from a frozen target worktree, record intentionally deferred verification without repeating covered tests, hand work to another Agent, recover a blank-session next action, initialize or inspect `.owlcoda/runkit/`, plan bounded work with engine pins and leases, capture exact verification evidence, close an execution, produce a ready-for-commit receipt, or perform a read-only foreign-project shadow without widening Git or release authority.
 ---
 
 # OwlCoda RunKit
 
 Use the bundled deterministic Core to coordinate work through project artifacts rather than chat memory. Require Node.js 20 or later.
 
-Bundled release: standalone `owlrunkit@0.18.3`, bundling Contract `0.2` and
-Core `0.18.3` with manifest
-`sha256:cc7a923649592084150828f24f0d90f280af92514bafdbc4999f4015b3175d5c`.
-The published `owlrunkit@0.18.2` package is the trusted prior for rollback;
+Bundled release: standalone `owlrunkit@0.19.0`, bundling Contract `0.2` and
+Core `0.19.0` with manifest
+`sha256:803fb7de76a6ccffd1ec5e0164eeff15a580209fa6dfbef9e561345ab616b403`.
+The published `owlrunkit@0.18.4` package is the trusted prior for rollback;
 verify exact official npm registry provenance before adoption. The immutable
 rollback baseline for the prior release is `owlrunkit@0.16.1` with SHA-1
 `1bc119d24879bdd8959af16b364af55ca9d61ea9` and canonical npm tarball URL
@@ -26,7 +26,9 @@ the user to operate the CLI when the request and authority are already clear.
 
 The normal mainline is `init`, `assign`, `checkpoint`, `handoff`,
 `reject-and-return`, `decision`, `verification`, `integrate`, `status`, and
-`takeover`. After a project
+`takeover`. External source lanes use `target-snapshot` and `import-delivery`
+without installing RunKit or writing control artifacts into the frozen target.
+After a project
 derives completed, `successor` is the bounded lifecycle step. `init` accepts one project definition
 JSON; subsequent actions use direct typed flags. Normal users do not author
 request, event, receipt, or hash JSON. The legacy `--request <event.json>` form
@@ -45,7 +47,13 @@ remains available for existing automation and compatibility.
   reject-and-return` event. Preserve the failed checkpoint, bind the review
   evidence, create the rework assignment, and never rewrite failure as active.
 - "Continue/take over": build `project takeover` for the current Agent and use
-  its bounded next action and evidence refs.
+  its bounded next action and evidence refs. Report stale handoffs as obsolete,
+  and use the projected current external candidate plus its predecessor delta
+  instead of reconstructing source identity from chat.
+- "Import Luna/Cursor delivery": capture the target entry before source work
+  when possible, then verify and import its exact DeliveryPacket. Without the
+  entry snapshot, report `candidate_only`; never claim an entry-to-candidate
+  overlay proof that was not captured.
 - "Start the next project": only after status derives `overall=completed`,
   invoke `project successor` with a different V1 definition and explicit
   transition ID. Preserve the content-addressed archive and journal; never
@@ -70,39 +78,53 @@ Completion requires evidence. If an assignment binds an execution, derive its
 state from the closeout and lease rather than trusting the Agent's claim.
 
 ```bash
-owlrunkit project init --workspace "$PROJECT_ROOT" --definition "$PROJECT_JSON"
-owlrunkit project assign --workspace "$PROJECT_ROOT" --assignment-id "$ASSIGNMENT_ID" \
+npx --no-install owlrunkit project init --workspace "$PROJECT_ROOT" --definition "$PROJECT_JSON"
+npx --no-install owlrunkit project assign --workspace "$PROJECT_ROOT" --assignment-id "$ASSIGNMENT_ID" \
   --at "$OCCURRED_AT" --work-item "$WORK_ITEM_ID" --agent "$AGENT_ID"
-owlrunkit project checkpoint --workspace "$PROJECT_ROOT" --checkpoint-id "$CHECKPOINT_ID" \
+npx --no-install owlrunkit project checkpoint --workspace "$PROJECT_ROOT" --checkpoint-id "$CHECKPOINT_ID" \
   --at "$OCCURRED_AT" --assignment-id "$ASSIGNMENT_ID" --work-item "$WORK_ITEM_ID" \
   --state active --summary "$SUMMARY" --next "$NEXT_ACTION"
-owlrunkit project handoff --workspace "$PROJECT_ROOT" --handoff-id "$HANDOFF_ID" \
+npx --no-install owlrunkit project handoff --workspace "$PROJECT_ROOT" --handoff-id "$HANDOFF_ID" \
   --at "$OCCURRED_AT" --assignment-id "$ASSIGNMENT_ID" --work-item "$WORK_ITEM_ID" \
   --from-agent "$AGENT_ID" --to-agent "$SUCCESSOR_AGENT_ID" \
   --summary "$SUMMARY" --next "$NEXT_ACTION" --evidence "$EVIDENCE_REF"
-owlrunkit project reject-and-return --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit project reject-and-return --workspace "$PROJECT_ROOT" \
   --rework-id "$REWORK_ID" --at "$OCCURRED_AT" --work-item "$WORK_ITEM_ID" \
   --reviewer-agent "$REVIEWER_AGENT_ID" --to-agent "$AGENT_ID" \
   --reason "$REJECTION_REASON" --next "$NEXT_ACTION" --evidence "$EVIDENCE_REF"
-owlrunkit project decision --workspace "$PROJECT_ROOT" --resolve \
+npx --no-install owlrunkit project decision --workspace "$PROJECT_ROOT" --resolve \
   --decision-id "$DECISION_ID" --at "$OCCURRED_AT" --resolution "$RESOLUTION" \
   --rationale "$RATIONALE" --evidence "$EVIDENCE_REF"
-owlrunkit project verification --workspace "$PROJECT_ROOT" --defer \
+npx --no-install owlrunkit project verification --workspace "$PROJECT_ROOT" --defer \
   --verification-id "$VERIFICATION_ID" --at "$OCCURRED_AT" \
   --work-item "$WORK_ITEM_ID" --owner-agent "$AGENT_ID" \
   --check "$CHECK_ID" --reason "$REASON" --due-gate "$GATE_ID"
-owlrunkit project verification --workspace "$PROJECT_ROOT" --close \
+npx --no-install owlrunkit project verification --workspace "$PROJECT_ROOT" --close \
   --verification-id "$VERIFICATION_ID" --at "$OCCURRED_AT" \
   --disposition verified --summary "$SUMMARY" --evidence "$EVIDENCE_REF"
-owlrunkit project integrate --workspace "$PROJECT_ROOT" --gate "$GATE_ID" \
+npx --no-install owlrunkit project integrate --workspace "$PROJECT_ROOT" --gate "$GATE_ID" \
   --at "$OCCURRED_AT" --summary "$SUMMARY" --evidence "$EVIDENCE_REF"
-owlrunkit project status --workspace "$PROJECT_ROOT" --json
-owlrunkit project takeover --workspace "$PROJECT_ROOT" --agent "$AGENT_ID" --json
-owlrunkit project successor --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit project status --workspace "$PROJECT_ROOT" --json
+npx --no-install owlrunkit project takeover --workspace "$PROJECT_ROOT" --agent "$AGENT_ID" --json
+npx --no-install owlrunkit project successor --workspace "$PROJECT_ROOT" \
   --transition-id "$TRANSITION_ID" --at "$OCCURRED_AT" \
   --definition "$NEXT_PROJECT_JSON" --reason "$SUCCESSOR_REASON" --json
+npx --no-install owlrunkit project target-snapshot --workspace "$PROJECT_ROOT" \
+  --snapshot-id "$ENTRY_SNAPSHOT_ID" --at "$OCCURRED_AT" \
+  --work-item "$WORK_ITEM_ID" --target-workspace "$TARGET_WORKTREE" --apply
+npx --no-install owlrunkit project import-delivery --workspace "$PROJECT_ROOT" \
+  --delivery-id "$DELIVERY_ID" --at "$OCCURRED_AT" \
+  --assignment-id "$ASSIGNMENT_ID" --work-item "$WORK_ITEM_ID" \
+  --producer "$AGENT_ID" --target-workspace "$TARGET_WORKTREE" \
+  --packet "$DELIVERY_PACKET" --entry-snapshot "$ENTRY_SNAPSHOT_ID" --apply
 ```
 
+Controller and target must be disjoint real directories, and the target must
+be its exact Git worktree root. Import rechecks packet bytes and stable target
+identity, including both sides of renames. Do not use legacy raw project event
+requests to represent an external delivery; Event V3 is admitted only through
+`project import-delivery`.
+
 `project successor` accepts only a derived-completed active project and a
 different fresh V1 identity. It archives the exact prior bytes, never borrows
 old events, and resumes only an identical durable transition. Treat symlinks,
@@ -149,11 +171,16 @@ Git, release, deployment, or production action.
 4. Never include `.owlcoda/runkit/**` in delivery changed files, profile rules or lease owned paths.
 5. Read [references/contract-v0.2.md](references/contract-v0.2.md) for new executions and accepted closeouts. Read [references/contract-v0.1.md](references/contract-v0.1.md) only when interpreting historical v0.1 artifacts.
 
-Use the `owlrunkit` command for normal work. Its bootstrap resolves the
+Use `npx --no-install owlrunkit` for every command after the exact project
+dependency is installed. A bare `owlrunkit` may resolve an older global binary
+that cannot trust a newer project Core, so do not use it for project status,
+verification, or Project Driver actions. The package bootstrap resolves the
 project's exact locally bound OwlRunKit CLI before interpreting project control
-state. The currently installed package handles bootstrap-only operations such
+state. A current package may still be used for bootstrap-only operations such
 as initialization, exact adoption and Core-successor migration. Do not bypass
-that resolver with a global CLI or a copied Skill Core. Project-local
+the project-local entry with a global CLI or a copied/shared Skill Core. If a
+process-level Skill declares an older release, use this installed package Skill
+and report the activation drift; do not silently apply the old rules. Project-local
 delegation requires the canonical official npm tarball URL and SHA-512 SRI,
 an installed Core identity trusted by the current bootstrap, and the same Core
 identity in project Config; any mismatch fails closed.
@@ -163,11 +190,23 @@ below has no CLI entry.
 ## Initialize and inspect
 
 ```bash
-owlrunkit init --workspace "$PROJECT_ROOT"
-owlrunkit inspect --json --workspace "$PROJECT_ROOT"
-owlrunkit status --workspace "$PROJECT_ROOT"
+npx --no-install owlrunkit bootstrap --workspace "$PROJECT_ROOT" \
+  --exact owlrunkit@0.19.0 --dry-run
+npx --no-install owlrunkit bootstrap --workspace "$PROJECT_ROOT" \
+  --exact owlrunkit@0.19.0 --apply
+npx --no-install owlrunkit init --workspace "$PROJECT_ROOT"
+npx --no-install owlrunkit inspect --json --workspace "$PROJECT_ROOT"
+npx --no-install owlrunkit status --workspace "$PROJECT_ROOT"
 ```
 
+Prefer `bootstrap` for a new standalone-package consumer or an eligible idle
+project upgrade. Its dry-run is zero-write; apply rechecks the exact local
+package and official registry binding under a project-owned lock, then performs
+Core initialization, profile reconcile/validation, adoption and doctor
+readback. A later-step failure restores the governed preimage and leaves a
+failure receipt. Shared Skill inspection is diagnostic only; bootstrap never
+silently changes a user or fleet Skill.
+
 Initialization may write only `.owlcoda/runkit/`. It must not modify business source.
 It atomically upgrades a known ConfigV1 project or refreshes an older v0.2
 Core patch identity and writes a local migration receipt with
@@ -204,40 +243,53 @@ repair surfaces. OwlRunKit uses its own version lifecycle: the standalone
 For one low-risk exact command, use the Core v0.13 Quick path:
 
 ```bash
-owlrunkit quick-verify \
+npx --no-install owlrunkit quick-verify \
   --workspace "$PROJECT_ROOT" -- "$ABSOLUTE_EXECUTABLE" "arg-one"
 
-owlrunkit quick-verify \
+npx --no-install owlrunkit quick-verify \
   --workspace "$PROJECT_ROOT" --stdin-file "$SCRIPT_FILE" --attest \
   -- ssh test-host /bin/sh
 
-owlrunkit quick-attest \
+npx --no-install owlrunkit quick-attest \
   --workspace "$PROJECT_ROOT" --receipt "$QUICK_RECEIPT"
 
-owlrunkit quick-metrics \
-  --workspace "$PROJECT_ROOT" --local
+npx --no-install owlrunkit quick-metrics \
+  --workspace "$PROJECT_ROOT"
+
+npx --no-install owlrunkit quick-metrics \
+  --workspace "$PROJECT_ROOT" --verbose
 
-owlrunkit quick-metrics \
-  --workspace "$PROJECT_ROOT" --local --verbose
+npx --no-install owlrunkit efficiency status \
+  --workspace "$PROJECT_ROOT"
 ```
 
 `--stdin-file` copies and SHA-256-binds at most 1 MiB of regular non-symlink
 input before supplying the stored bytes as exact stdin. `--attest` returns the
 separate attestation object in the same high-level invocation. Quick receipts
 remain `captured_verification`, not Formal acceptance. Metrics are aggregate by
-default; `--verbose` adds receipt paths. Both forms are local-only and perform
-no telemetry or network request.
+default; `--verbose` adds receipt paths. Metrics are always local-only and
+perform no telemetry or network request; the former `--local` flag remains an
+accepted compatibility no-op.
+
+Quick's default human result leads with pass/fail and explicitly separates
+candidate identity, workspace attestation, verification, product/UX,
+deployment/runtime, and production/business evidence. A lower layer never
+promotes a higher layer. Use `efficiency record` only for an observed workflow
+start, confirmed false block, manual intervention, or named acceptable result;
+use `efficiency status` to inspect local repeated commands, handoff wait, and
+time-to-result. These metrics are project-local, perform no telemetry, exclude
+receipt count as a value signal, and grant no additional authority.
 
 Core v0.13 also adds deterministic repair and exact-byte offline transport:
 
 ```bash
-owlrunkit repair \
+npx --no-install owlrunkit repair \
   --workspace "$PROJECT_ROOT" --run-id "$RUN_ID"
 
-owlrunkit offline-export \
+npx --no-install owlrunkit offline-export \
   --workspace "$PROJECT_ROOT" --receipt "$RECEIPT" --output "$BUNDLE"
 
-owlrunkit offline-import \
+npx --no-install owlrunkit offline-import \
   --workspace "$PROJECT_ROOT" --bundle "$BUNDLE" --store "$OFFLINE_STORE"
 ```
 
@@ -272,7 +324,7 @@ loaded into the current session and grants no update authority.
 The install manifest binds every Skill, Core, contract and attest file.
 Unknown installed drift fails closed instead of being overwritten. A changed
 process-wide Skill requires a registered complete fleet. Register each coverage
-root once with `owlrunkit fleet register-root --fleet-root <absolute-root>`.
+root once with `npx --no-install owlrunkit fleet register-root --fleet-root <absolute-root>`.
 Inspect and correct a broad or stale registry through the official
 `fleet inspect-registry` and evidence-bound `fleet replace-registry` workflow;
 never hand-edit the registry. Replacement is dry-runnable and append-only
@@ -330,11 +382,11 @@ made inactive 0.13/0.14 ConfigV2 upgrades and exact rollback safe. Historical
 projects, profiles, executions, packets, receipts and closeouts remain readable.
 
 ```bash
-owlrunkit doctor --workspace "$PROJECT_ROOT"
-owlrunkit profiles detect --workspace "$PROJECT_ROOT" --dry-run
-owlrunkit profiles validate --workspace "$PROJECT_ROOT"
-owlrunkit profiles impact --workspace "$PROJECT_ROOT" --changed path/to/file
-owlrunkit adopt --workspace "$PROJECT_ROOT" --exact owlrunkit@0.15.1
+npx --no-install owlrunkit doctor --workspace "$PROJECT_ROOT"
+npx --no-install owlrunkit profiles detect --workspace "$PROJECT_ROOT" --dry-run
+npx --no-install owlrunkit profiles validate --workspace "$PROJECT_ROOT"
+npx --no-install owlrunkit profiles impact --workspace "$PROJECT_ROOT" --changed path/to/file
+npx --no-install owlrunkit adopt --workspace "$PROJECT_ROOT" --exact owlrunkit@0.15.1
 ```
 
 Exact adoption requires the official npm registry `shasum`, `integrity` and
@@ -353,15 +405,15 @@ authority.
 Core 0.16 makes the safe path the ordinary path without weakening its gates:
 
 ```bash
-owlrunkit fleet register-root --fleet-root /absolute/declared-projects
-owlrunkit fleet discover
-owlrunkit profiles detect --workspace "$PROJECT_ROOT" --dry-run
-owlrunkit profiles detect --workspace "$PROJECT_ROOT" --apply
-owlrunkit profiles reconcile --workspace "$PROJECT_ROOT" --dry-run
-owlrunkit mode recommend --workspace "$PROJECT_ROOT"
-owlrunkit assurance route --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit fleet register-root --fleet-root /absolute/declared-projects
+npx --no-install owlrunkit fleet discover
+npx --no-install owlrunkit profiles detect --workspace "$PROJECT_ROOT" --dry-run
+npx --no-install owlrunkit profiles detect --workspace "$PROJECT_ROOT" --apply
+npx --no-install owlrunkit profiles reconcile --workspace "$PROJECT_ROOT" --dry-run
+npx --no-install owlrunkit mode recommend --workspace "$PROJECT_ROOT"
+npx --no-install owlrunkit assurance route --workspace "$PROJECT_ROOT" \
   --request "$ASSURANCE_REQUEST_JSON"
-owlrunkit status --workspace "$PROJECT_ROOT"
+npx --no-install owlrunkit status --workspace "$PROJECT_ROOT"
 ```
 
 Fleet discovery is complete only within all registered coverage roots, an
@@ -372,14 +424,18 @@ adopts only unambiguous high-confidence, project-local regular-file launchers;
 `profiles reconcile` previews or atomically updates an existing profile set
 with a rollback receipt. Low-confidence candidates remain untouched. Lock-bound npm and exact
 project-installed pnpm, Yarn, and Bun launchers are supported; PATH-only
-package managers remain review-required. `assurance route`
+package managers remain review-required. A V2 dry-run reports
+`profiles_insufficient` when real Git-visible source paths remain uncovered or
+there is no actionable primary. Use its `minimalSuggestedProfile` and copyable
+`repairCommands`; do not invent absent conventional directories as blockers.
+`assurance route`
 deterministically selects No RunKit, Quick Verification or Formal Delivery
 from explicit risk facts and fails closed to Formal when required facts are
 missing.
 
 ## Choose continuity and assurance independently
 
-Start with `owlrunkit mode recommend --workspace "$PROJECT_ROOT"`. The five
+Start with `npx --no-install owlrunkit mode recommend --workspace "$PROJECT_ROOT"`. The five
 user-facing modes are presets over two independent dimensions:
 
 - continuity: `none | project_driver`
@@ -419,17 +475,17 @@ incremental checks, then finish once:
 ```bash
 NODE_BIN="$(node -p 'process.execPath')"
 
-owlrunkit formal start \
+npx --no-install owlrunkit formal start \
   --workspace "$PROJECT_ROOT" --run-id "$RUN_ID" --goal "$GOAL_JSON" \
   --work-item "$WORK_ITEM_ID" \
   --owned-path "src/exact-file.ts" --owned-path "tests/feature/**"
 
-owlrunkit formal check \
+npx --no-install owlrunkit formal check \
   --workspace "$PROJECT_ROOT" --run-id "$RUN_ID" \
   --from-lease "$WORK_ITEM_ID" --check-id "$CHECK_ID" \
   --cwd "." -- "$NODE_BIN" --check "src/exact-file.ts"
 
-owlrunkit formal finish \
+npx --no-install owlrunkit formal finish \
   --workspace "$PROJECT_ROOT" --run-id "$RUN_ID" --decision accepted
 ```
 
@@ -446,7 +502,7 @@ Formal accepted evidence admits the built-in Node syntax check and real project
 commands executed through a Verification Envelope:
 
 ```bash
-owlrunkit formal check \
+npx --no-install owlrunkit formal check \
   --workspace "$PROJECT_ROOT" --run-id "$RUN_ID" \
   --work-item "$WORK_ITEM_ID" --check-id "$CHECK_ID" \
   --envelope "$VERIFICATION_ENVELOPE_JSON"
@@ -467,12 +523,12 @@ Freeze, verify and materialize an immutable candidate without requiring a Git
 commit:
 
 ```bash
-owlrunkit candidate freeze --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit candidate freeze --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" --from-lease "$WORK_ITEM_ID" \
   --candidate-id "$CANDIDATE_ID"
-owlrunkit candidate verify --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit candidate verify --workspace "$PROJECT_ROOT" \
   --candidate "$SOURCE_CANDIDATE"
-owlrunkit candidate materialize --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit candidate materialize --workspace "$PROJECT_ROOT" \
   --candidate "$SOURCE_CANDIDATE" \
   --target-workspace "$CLEAN_WORKSPACE"
 ```
@@ -490,10 +546,10 @@ authority.
 Use a frozen SourceCandidateV2 and the registered fleet boundary:
 
 ```bash
-owlrunkit core-successor plan --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit core-successor plan --workspace "$PROJECT_ROOT" \
   --plan-id "$PLAN_ID" --run-id "$RUN_ID" \
   --from-lease "$WORK_ITEM_ID" --candidate-id "$CANDIDATE_ID"
-owlrunkit core-successor apply --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit core-successor apply --workspace "$PROJECT_ROOT" \
   --plan "$SUCCESSOR_PLAN" --owner-authority "$OWNER_AUTHORITY" \
   --receipt-id "$RECEIPT_ID"
 ```
@@ -510,11 +566,11 @@ execute. No execution result is silently inherited.
 Separate build acceptance from permission to mutate a machine:
 
 ```bash
-owlrunkit deployment prepare --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit deployment prepare --workspace "$PROJECT_ROOT" \
   --run-id "$PREPARE_RUN_ID" --artifact "$ARTIFACT" \
   --owner-decision "$OWNER_DECISION" \
   --media-type application/gzip --output "$PREPARE_RECEIPT"
-owlrunkit deployment execute --workspace "$PROJECT_ROOT" \
+npx --no-install owlrunkit deployment execute --workspace "$PROJECT_ROOT" \
   --prepare "$PREPARE_RECEIPT" \
   --owner-decision "$OWNER_DECISION" \
   --owner-authority "$OWNER_AUTHORITY" \
@@ -567,7 +623,7 @@ and protocol contract, not an out-of-the-box privileged VM deployment.
 Acquire one bounded writer lease with repeated explicit paths:
 
 ```bash
-owlrunkit lease acquire \
+npx --no-install owlrunkit lease acquire \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --work-item "$WORK_ITEM_ID" \
@@ -578,10 +634,10 @@ owlrunkit lease acquire \
 Inspect or release it through the Core instead of editing lease JSON:
 
 ```bash
-owlrunkit lease inspect \
+npx --no-install owlrunkit lease inspect \
   --workspace "$PROJECT_ROOT" --run-id "$RUN_ID"
 
-owlrunkit lease release \
+npx --no-install owlrunkit lease release \
   --workspace "$PROJECT_ROOT" --run-id "$RUN_ID" \
   --work-item "$WORK_ITEM_ID"
 ```
@@ -589,7 +645,7 @@ owlrunkit lease release \
 After implementation, derive the bounded DeliveryPacket from the active lease:
 
 ```bash
-owlrunkit delivery create \
+npx --no-install owlrunkit delivery create \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --from-lease "$WORK_ITEM_ID" \
@@ -625,7 +681,7 @@ If any path is uncovered, use the declared full profile. Do not interpret an emp
 Before reusing evidence, prepare `OwlCodaRunKitCoverageAdoptRequestV1` and run:
 
 ```bash
-owlrunkit coverage-adopt \
+npx --no-install owlrunkit coverage-adopt \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --request "$COVERAGE_ADOPT_REQUEST_JSON"
@@ -645,7 +701,7 @@ After inspecting the source execution, prepare `OwlCodaRunKitResumeRequestV1`
 and run:
 
 ```bash
-owlrunkit resume \
+npx --no-install owlrunkit resume \
   --workspace "$PROJECT_ROOT" \
   --run-id "$SOURCE_RUN_ID" \
   --request "$RESUME_REQUEST_JSON"
@@ -675,7 +731,7 @@ After freezing a DeliveryPacket, prepare
 SHA-256, then run:
 
 ```bash
-owlrunkit verify-plan \
+npx --no-install owlrunkit verify-plan \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --request "$VERIFY_PLAN_REQUEST_JSON"
@@ -706,7 +762,7 @@ For a model workload, first bind the remaining work to a hashed
 `OwlCodaRunKitResourcePreflightRequestV1` and run:
 
 ```bash
-owlrunkit resource-preflight \
+npx --no-install owlrunkit resource-preflight \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --request "$RESOURCE_PREFLIGHT_REQUEST_JSON"
@@ -751,7 +807,7 @@ labels are not acceptance evidence.
 Prepare `OwlCodaRunKitSnapshotRequestV1`, then run:
 
 ```bash
-owlrunkit snapshot \
+npx --no-install owlrunkit snapshot \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --request "$SNAPSHOT_REQUEST_JSON"
@@ -773,7 +829,7 @@ and after the command.
 Prepare `OwlCodaRunKitVisualSmokeRequestV1`, then run:
 
 ```bash
-owlrunkit visual-smoke \
+npx --no-install owlrunkit visual-smoke \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --request "$VISUAL_SMOKE_REQUEST_JSON"
@@ -793,7 +849,7 @@ After every required snapshot passed and the DeliveryPacket is still fresh,
 prepare `OwlCodaRunKitFinalizeRequestV1` and run:
 
 ```bash
-owlrunkit finalize \
+npx --no-install owlrunkit finalize \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --request "$FINALIZE_REQUEST_JSON"
@@ -810,7 +866,7 @@ this command is available.
 Close only after consuming the current stage receipt and making an explicit controller decision:
 
 ```bash
-owlrunkit closeout \
+npx --no-install owlrunkit closeout \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --decision accepted \
@@ -831,7 +887,7 @@ After an accepted closeout, capture fresh project-root snapshots and start from
 `sourceCandidatePath` or legacy `deliveryPacketPath`, then run:
 
 ```bash
-owlrunkit ready-for-commit \
+npx --no-install owlrunkit ready-for-commit \
   --workspace "$PROJECT_ROOT" \
   --run-id "$RUN_ID" \
   --request "$READY_REQUEST_JSON"

package/integrations/codex/skills/owlcoda-runkit/assets/templates/config.json

@@ -2,9 +2,9 @@
   "schemaVersion": "OwlCodaRunKitConfigV2",
   "core": {
     "contractVersion": "0.2",
-    "coreVersion": "0.18.3",
-    "coreManifestSha256": "sha256:cc7a923649592084150828f24f0d90f280af92514bafdbc4999f4015b3175d5c",
-    "coreSourceRef": "artifact:sha256:cc7a923649592084150828f24f0d90f280af92514bafdbc4999f4015b3175d5c"
+    "coreVersion": "0.19.0",
+    "coreManifestSha256": "sha256:803fb7de76a6ccffd1ec5e0164eeff15a580209fa6dfbef9e561345ab616b403",
+    "coreSourceRef": "artifact:sha256:803fb7de76a6ccffd1ec5e0164eeff15a580209fa6dfbef9e561345ab616b403"
   },
   "authorizationPolicy": "external_explicit_authority_required"
 }

package/integrations/codex/skills/owlcoda-runkit/references/contract-v0.2.md

@@ -478,6 +478,44 @@ bytes as exact stdin, and binds their workspace-relative path, byte size, and
 SHA-256. V1 remains the no-stdin receipt and stays readable. Neither schema
 binds undeclared ignored outputs or grants any authority.
 
+Core `0.19.0` adds `OwlCodaQuickVerificationReceiptV3` for explicit isolation,
+dependency-environment binding, and ignored-path binding. `--isolate` runs a
+Git-visible candidate copy in a disposable consumer and records consumer source
+versus discarded ignored-artifact deltas. `--dependency-root` binds an existing
+dependency environment without installing it. `--bind-ignored` is the only
+operation that emits `OwlCodaWorkspaceSnapshotV2` and claims those selected
+ignored bytes. Workspace source, selected ignored paths, dependency metadata,
+and consumer source mutations are distinct fail-closed facts. Disposable
+ignored artifacts do not mutate or become evidence for the source checkout.
+
+Core `0.19.0` also adds `OwlCodaQuickEvidenceReusePlanV1` and
+`OwlCodaQuickEvidenceReuseReceiptV1`. A reuse plan verifies one earlier strict,
+passed, source-unchanged Quick receipt from the same workspace, derives the
+current candidate delta from Git objects plus both snapshots, and refuses any
+path outside literal caller allowances before command execution. Byte-identical
+dirty content later committed to Git remains unchanged evidence. Apply runs one
+exact delta gate through Quick and records both receipt hashes; it grants no
+authority and does not satisfy Formal or higher acceptance layers.
+
+Quick output and Verification Envelope stdout/stderr evidence use a project-
+global create-only SHA-256 object store. Existing Quick receipt-local material
+paths remain valid; no historical tree is rewritten or deleted. Object reuse is
+storage/provenance behavior, never additional assurance.
+
+Core `0.19.0` can additionally bind external transition truth without changing
+Quick Receipt V1/V2/V3. `quick-verify --capture-runtime` gives the exact child
+an initially absent receipt-local `OWLRUNKIT_RUNTIME_REPORT` path. A strict
+`OwlCodaRunKitExternalRuntimeReportV1` there is projected into a separately
+hashed `OwlCodaRunKitRuntimeTransitionReceiptV1` that binds the Quick receipt
+bytes, report bytes, derived workspace mutation, command-reported external
+mutation, ordered stages, rollback outcome, final external state, and recovery
+artifact `capturedBefore` / `restoresTo` semantics. A failed stage followed by
+a successful rollback remains a failed transition with useful evidence; it is
+not promoted to Quick, Formal, deployment, or business acceptance. Missing or
+inconsistent runtime truth fails closed while preserving the base Quick
+receipt. `quick-runtime-verify` rechecks stored bytes and projection integrity;
+it does not independently re-observe the external runtime or grant authority.
+
 Deterministic `repair` persists a reviewable plan before replay, reuses only
 fresh receipt-bound coverage, replays pending exact argv, and appends
 replacement lineage. Failed and orphaned attempts remain inspectable. Repair
@@ -539,6 +577,25 @@ stale Packet, invalid evidence or engine drift must still follow the Core's
 foreign-project authority. Signing, keys, external anchors and GitHub Actions
 remain deferred.
 
+## Core v0.19 atomic onboarding
+
+Core `0.19.0` keeps Contract `0.2` and all existing authority boundaries. The
+versioned `OwlCodaRunKitOnboardingBootstrapPlanV1` dry-run binds the exact
+project-local package and official registry release, checks upgrade safety,
+previews profile reconciliation, reports shared Skill drift, and performs zero
+writes. `bootstrap --apply` rechecks that plan under a project-owned lock, then
+runs Core initialization, profile reconcile/validation, exact adoption and
+doctor readback as one high-level transaction.
+
+The transaction governs RunKit config, profiles, their migration/apply
+receipts, and adoption evidence. A failure after mutation restores their exact
+preimage before writing an append-only
+`OwlCodaRunKitOnboardingBootstrapReceiptV1`. A concurrent bootstrap, active or
+invalid project lifecycle, ambiguous project tool, exact-package mismatch, or
+registry mismatch fails closed. The command does not modify package files,
+source, Git, deployments, production, business state, or shared Skills, and it
+never grants those authorities.
+
 ## Deferred work
 
 Contract v0.2 still does not add a built-in browser or Electron driver, Plugin,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "owlrunkit",
-  "version": "0.18.3",
+  "version": "0.19.0",
   "description": "Project Driver for cross-Agent project continuity and progression; execution, evidence, verification, and recovery are the underlying trust substrate.",
   "license": "GPL-3.0-or-later",
   "author": "yeemio <owlclaw@163.com>",

package/packages/attest/README.md

@@ -1,15 +1,18 @@
 # `@owlcoda/attest`
 
 `@owlcoda/attest` is OwlCoda RunKit's unsigned, read-only verification plane.
-It verifies Quick Receipt V1 and V2 artifacts, Formal public verification
+It verifies Quick Receipt V1, V2, and V3 artifacts, Formal public verification
 bundles, and local `OwlCodaAttestationRefV1` references without importing the
 private mutation Core. V2 adds one hash-bound stdin material while preserving
-the same unsigned, authorization-free assurance boundary.
+the same unsigned, authorization-free assurance boundary. V3 adds a disposable
+consumer record, optional dependency-environment binding, and explicit ignored
+workspace bindings while leaving V1/V2 unchanged.
 
 The package is an unsigned V1 distribution candidate. It is intentionally
 marked private until naming clearance and release authorization exist.
-`consumer-surface-v1.json` freezes the complete command, schema, reference,
-issue-code, compatibility, privacy, metrics, and unsupported-claim covenant.
+`consumer-surface-v1.json` remains the byte-compatible predecessor contract.
+`consumer-surface-v2.json` adds Receipt V3, Snapshot V2, isolated Quick, and
+explicit ignored-path issue codes without reinterpreting V1.
 
 ## CLI
 
@@ -58,7 +61,8 @@ in the repository documentation under
 receipts but does not require N-1 to trust an unknown future Core identity.
 
 The packed candidate includes its GPL-3.0-or-later license, a package-level
-SPDX 2.3 SBOM, and the frozen Consumer Surface V1 manifest. Local metrics never
+SPDX 2.3 SBOM, the frozen Consumer Surface V1 manifest, and the additive V2
+surface. Local metrics never
 send telemetry and are not a performance target. They describe local receipt
 state; separate consumption must be evidenced by a caller-owned attestation
 result.