docs: align execution handoff with Glas

.custodian-brief.md

@@ -2,7 +2,7 @@
 # Custodian Brief — kaizen-agentic
 
 **Domain:** infotech
-**Last synced:** 2026-08-20 20:43 UTC
+**Last synced:** 2026-08-21 06:29 UTC
 **State Hub:** http://127.0.0.1:8000 *(adjust if running on a remote machine)*
 
 ## Active Workstreams

README.md

@@ -117,7 +117,7 @@ Feedback: `kaizen-agentic feedback` · [docs/FEEDBACK.md](docs/FEEDBACK.md)
 - **Agency Framework**: Project-scoped agent memory + Coach meta-agent for cross-agent synthesis
 - **Metrics and Improvement Loop**: Session evidence, optimizer recommendations, and optional artifact publication
 - **Roles and Engagements**: Versioned role packages, client-bound engagement records, lifecycle checklists, and custody controls
-- **Scheduled Preparation**: Repo-local schedules and offline orientation bundles for activity-core/rein-aharness execution
+- **Scheduled Preparation**: Repo-local schedules and offline orientation bundles for activity-core/Glas execution
 - **CLI Tool**: Agent, memory, metrics, protocol, engagement, and schedule commands (`kaizen-agentic`)
 - **Project Templates**: Pre-configured setups for different project types
 - **Runtime-neutral Contracts**: Instruction and preparation surfaces usable by governed coding-agent harnesses
@@ -159,9 +159,10 @@ Feedback: `kaizen-agentic feedback` · [docs/FEEDBACK.md](docs/FEEDBACK.md)
 
 Kaizen Agentic defines agent craft, validates repo-local schedules, and prepares
 offline execution bundles. In the current unattended path, **activity-core**
-creates durable, claimable `ops_run` work and **rein-aharness** resolves policy,
+creates durable, claimable `ops_run` work. A Kaizen blueprint or agent instance
-credentials, runtime profile, model, and sandbox before executing the bounded
+hands a versioned `harness_profile_ref` to **glas-harness**, which resolves the
-agent session. State Hub supplies roster and coordination evidence.
+concrete rein, model route, sandbox, tool policy, and limits and returns one
+evidence envelope. State Hub supplies roster and coordination evidence.
 
 Manual execution remains supported. Kaizen Agentic itself does not own cron,
 the durable work queue, inference, credentials, or runtime authorization. See

SCOPE.md

@@ -24,9 +24,11 @@ and optimization proposes the next safe refinement.
 KaizenAgentic **declares and prepares** agent work. It does not own the clock,
 credentials, policy enforcement, or LLM session runtime. In the current
 ecosystem, activity-core decides when/what/where and emits claimable work;
-rein-aharness binds a repo-owned instance to a Kaizen blueprint and executes a
+a repo-owned instance binds a Kaizen blueprint to a versioned
-bounded session. This separation is the implementation of the strategic
+`harness_profile_ref`; glas-harness resolves the concrete execution
-boundary in `INTENT.md`, not an unfinished runtime hidden in this repository.
+constellation; and the selected rein executes a bounded session. This
+separation is the implementation of the strategic boundary in `INTENT.md`, not
+an unfinished runtime hidden in this repository.
 
 ---
 
@@ -52,9 +54,9 @@ boundary in `INTENT.md`, not an unfinished runtime hidden in this repository.
   packages, staffed-role lifecycle, bound agent definitions, ramp-up/down,
   client-custodied memory, engagement metrics, and reference pilot artifacts
 - **Interoperability contracts and reference patterns** for activity-core,
-  rein-aharness, role-engine, agentic-resources, human-resources, fin-hub,
+  glas-harness, the selected reins, role-engine, agentic-resources,
-  resource-control, llm-connect, State Hub, artifact-store, Helix Forge, and
+  human-resources, fin-hub, resource-control, llm-connect, State Hub,
-  Custodian discovery
+  artifact-store, Helix Forge, and Custodian discovery
 - **ADRs, workplans, tutorials, and integration handoffs** that make these
   conventions human-readable and machine-consumable
 
@@ -94,8 +96,8 @@ boundary in `INTENT.md`, not an unfinished runtime hidden in this repository.
   to an executable agent blueprint
 - Translating leadership-owned goals and budget-constrained assignments into
   measurable blueprint, protocol, schedule, and evidence contracts
-- Integrating Kaizen work with activity-core, rein-aharness, State Hub, Helix
+- Integrating Kaizen work with activity-core, glas-harness, a selected rein,
-  Forge, or artifact-store
+  State Hub, Helix Forge, or artifact-store
 - Understanding the KaizenAgentic product and agency operating model
 
 ---
@@ -103,8 +105,8 @@ boundary in `INTENT.md`, not an unfinished runtime hidden in this repository.
 ## Not Relevant When
 
 - A task needs only ad-hoc automation with no reusable agent or improvement loop
-- The need is purely schedule/runtime operation; use activity-core and the
+- The need is purely schedule/runtime operation; use activity-core,
-  selected rein/harness
+  glas-harness, and the selected rein
 - The need is organizational role definition only; use role-engine
 - The need is agent inventory, assignment, capacity planning, portfolio review,
   or retirement; agentic-resources is the intended owner once its boundary is
@@ -127,8 +129,9 @@ boundary in `INTENT.md`, not an unfinished runtime hidden in this repository.
 - **Improvement loop:** metrics storage, summary, optimization, correlation,
   publishing, and `kaizen.metrics.recorded` event emission are implemented
 - **Automation contract:** schedule manifests and offline prepare bundles are
-  implemented; activity-core has a durable `ops_run` queue and rein-aharness has
+  implemented; activity-core has a durable `ops_run` queue; glas-harness
-  a live claim loop plus a generic `agent-session` approach
+  contract 1.0 resolves versioned execution profiles and is proven across two
+  rein/model constellations; rein-local intake remains where already implemented
 - **Agency model:** one supplier Role package and one staffed host-operator pilot
   prove the file conventions; broader role-catalog integration is not yet
   standardized
@@ -153,7 +156,8 @@ boundary in `INTENT.md`, not an unfinished runtime hidden in this repository.
 | Monetary budgets, commitments, burn, runway, and AI-plan booked cost | `fin-hub` | Consume provenance-bearing constraints; publish goal/assignment/duty attribution evidence |
 | Resource demand, capacity, allocation, utilization, and technical economics | `resource-control` | Consume capacity/cost evidence; do not duplicate resource or procurement authority |
 | Scheduling and claimable work | `activity-core` | Supply manifests, definitions, prepare commands, and event contracts |
-| Governed unattended execution | `rein-aharness` (selected rein) | Supply blueprint/orientation bundle; receive metrics and improvement evidence |
+| Governed execution abstraction | `glas-harness` | Supply versioned `harness_profile_ref` plus blueprint/orientation and organizational references; consume normalized evidence |
+| Inner execution loop and backend policy | Glas-selected rein (`rein-aharness`, `rein-openweights`, …) | Remain rein-neutral; receive metrics and improvement evidence |
 | Model/provider abstraction | `llm-connect` | Runtime dependency outside this repository |
 | Credentials and authorization | OpenBao/credential routing, identity and policy systems | Reference named routes/profiles only; never hold grants or secrets |
 | Durable coordination evidence | Custodian State Hub | Publish/index progress and workplan state; do not own hub code |
@@ -245,7 +249,7 @@ keywords: [metrics, feedback, optimization, evidence, events]
 type: integration
 title: Scheduled-agent declaration and preparation
 description: Validate repo-local schedules and assemble offline orientation bundles for activity-core-triggered, harness-executed runs.
-keywords: [schedule, activity-core, rein-aharness, preparation, automation]
+keywords: [schedule, activity-core, glas-harness, preparation, automation]
 ```
 
 ```capability
@@ -264,5 +268,6 @@ keywords: [kaizen, intent, template, optimization, operating-model]
   `make agents-sync-package`.
 - Agent definitions still use less structure than the full
   `wiki/KaizenAgentTemplate.md`; schema conformance is a maturity gap.
-- `.kaizen/schedule.yml` base fields are owned here. rein-aharness adds runtime
+- `.kaizen/schedule.yml` base fields are owned here. Consumer-owned agent
-  policy extensions and validates them separately.
+  instances carry a versioned `harness_profile_ref`; Glas and the selected rein
+  validate runtime policy separately.

WORK-RECORDS.md

@@ -8,6 +8,7 @@
 
 | Kind | ID | Status | Lane | Source |
 | --- | --- | --- | --- | --- |
+| workplan | ADHOC-2026-08-21 | finished | — | workplans/ADHOC-2026-08-21.md |
 | workplan | KAIZEN-WP-0001 | finished | — | workplans/KAIZEN-WP-0001-community-engagement.md |
 | workplan | KAIZEN-WP-0002 | finished | — | workplans/KAIZEN-WP-0002-agency-framework.md |
 | workplan | KAIZEN-WP-0003 | finished | — | workplans/KAIZEN-WP-0003-measurement-loop.md |
@@ -22,6 +23,7 @@
 | workplan | KAIZEN-WP-0012 | finished | — | workplans/KAIZEN-WP-0012-scope-intent-role-automation-assessment.md |
 | workplan | KAIZEN-WP-0013 | finished | — | workplans/KAIZEN-WP-0013-agent-workforce-goal-budget-governance.md |
 | workplan | KAIZEN-WP-0014 | finished | — | workplans/KAIZEN-WP-0014-repository-consolidation.md |
+| task | ADHOC-2026-08-21-T01 | done | — | workplans/ADHOC-2026-08-21.md |
 | task | T01 | done | — | workplans/KAIZEN-WP-0003-measurement-loop.md |
 | task | T02 | done | — | workplans/KAIZEN-WP-0003-measurement-loop.md |
 | task | T03 | done | — | workplans/KAIZEN-WP-0003-measurement-loop.md |

docs/CLI_CHEAT_SHEET.md

@@ -120,9 +120,11 @@ kaizen-agentic schedule prepare optimization --format json
 ```
 
 activity-core fires the schedule and creates durable `ops_run` work per
-(repo, agent). rein-aharness claims that work and runs the governed
+(repo, agent). Rein-local intake or another caller receives that work and runs
-`agent-session`, including `schedule prepare`. kaizen-agentic does not run cron,
+`schedule prepare`; the blueprint/instance then hands an explicit versioned
-own the queue, resolve credentials, or invoke an LLM.
+`harness_profile_ref` to Glas for rein/model/sandbox/tool resolution.
+kaizen-agentic does not run cron, own the queue, resolve credentials, select a
+concrete rein, or invoke an LLM.
 
 ### Forward-deployed engagements (WP-0009 / DEC-FDA-001)
 ```bash

docs/INTEGRATION_PATTERNS.md

@@ -118,8 +118,10 @@ contract. Current unattended execution composes it as follows:
 |-------|-------|----------|
 | Schedule/event and eligibility | activity-core | Temporal definition/rule plus repo/agent context |
 | Durable work | activity-core | Unique idempotency key, `ops_run` state, lease, retry, result |
-| Claim and execution | rein-aharness | Claim loop, `agent-session` approach, policy/profile/runtime resolution |
+| Task intake | Caller or selected rein | Claim/receive bounded work; intake remains rein-local where implemented |
 | Agent orientation | kaizen-agentic | `schedule prepare <agent>` offline bundle |
+| Execution resolution | glas-harness | Explicit versioned `harness_profile_ref` → rein/model/sandbox/tools/limits; normalized evidence |
+| Inner execution | Glas-selected rein | Backend-specific bounded loop, policy enforcement, and credential use |
 | Coordination evidence | State Hub | Roster, workplan/progress references; never prompts or secrets |
 | Improvement evidence | kaizen-agentic / artifact-store | Project metrics and optional durable artifacts |
 
@@ -128,7 +130,8 @@ work. The automated path should keep prompts, secrets, and raw model output out
 of `ops_run` and State Hub records.
 
 **Boundary:** kaizen-agentic declares and prepares; activity-core schedules and
-queues; rein-aharness executes; State Hub owns the roster and coordination view.
+queues; glas-harness resolves the explicit execution profile; the selected rein
+executes; State Hub owns the roster and coordination view.
 
 ---
 

docs/adr/ADR-005-scheduled-agent-execution.md

@@ -55,18 +55,22 @@ specific durable implementation:
 activity-core Temporal schedule/event
   → rule evaluation and eligibility
   → unique, idempotent ops_run with lease/retry state
-  → rein-aharness claim loop
+  → rein-local intake or caller receives bounded work
-  → agent-session approach and governed runtime/profile resolution
   → kaizen-agentic schedule prepare <agent>
-  → bounded coding-agent session
+  → Glas ExecutionRequest(harness_profile_ref=...)
+  → versioned profile resolves rein + model + sandbox + tool policy + limits
+  → bounded session and normalized execution evidence
   → ops_run result + State Hub progress + .kaizen metrics/artifact evidence
 ```
 
-`activity-core` owns schedule and durable work state. `rein-aharness` owns
+`activity-core` owns schedule and durable work state. Task intake stays with
-claiming, execution policy, credential routing, runtime selection, and bounded
+the caller or rein where already implemented. A Kaizen blueprint or agent
-session operation. Kaizen Agentic continues to own declaration, agent craft,
+instance supplies an explicit, versioned `harness_profile_ref`; glas-harness
-preparation, and improvement evidence. A human-started coding-agent session
+resolves the concrete execution constellation and normalizes evidence; the
-remains a supported execution mode and reference smoke path.
+selected rein owns its inner loop, backend policy, and credential use. Kaizen
+Agentic continues to own declaration, agent craft, preparation, and improvement
+evidence. A human-started coding-agent session remains a supported execution
+mode and reference smoke path.
 
 ### 1. Schedule manifest — `.kaizen/schedule.yml`
 
@@ -149,9 +153,11 @@ kaizen-agentic schedule prepare <agent> [--target PATH] [--format markdown|json]
   rule evaluation, and `ops_run` state belong to activity-core; the roster query
   belongs to State Hub.
 - **No LLM invocation.** `prepare` produces a runner-agnostic bundle; a human or
-  rein-aharness-governed coding-agent session executes it.
+  Glas-selected rein executes it.
-- **No runtime authority or credentials.** Those remain with rein-aharness and
+- **No concrete rein/model selection.** Kaizen carries a versioned
-  the ecosystem credential/authorization owners.
+  `harness_profile_ref`; glas-harness resolves it without a hidden default.
+- **No runtime authority or credentials.** Those remain with Glas, the selected
+  rein, and the ecosystem credential/authorization owners.
 - **State-hub schema changes** (roster opt-in flag) are designed here but
   implemented in `the-custodian` (repo boundary).
 

docs/adr/ADR-007-forward-deployed-engagement-convention.md

@@ -132,16 +132,17 @@ should amend this ADR (or a successor), not re-open DEC-FDA-001 silently.
   `railiance-infra`; transfer requires explicit receiver acceptance before the
   supplier copy is removed or reduced to a pointer.
 - Manual preparation remains supported. Unattended host duties require an
-  engagement-specific activity-core definition and a rein-aharness execution
+  engagement-specific activity-core definition and an approved versioned Glas
-  profile in addition to the generic `ops_run`/`agent-session` path.
+  `harness_profile_ref` in addition to the generic `ops_run` intake path.
 - Kai is not a payment processor; EUR path remains future work
 
 ### Follow-ups
 
 - Complete the proposed pilot transfer to the client-owned `railiance-infra`
   evidence path and record its accepted commit
-- Register engagement-specific activity-core definitions and rein-aharness
+- Register engagement-specific activity-core definitions and a Glas execution
-  profile for daily host-operator preparation when unattended operation is desired
+  profile reference for daily host-operator preparation when unattended
+  operation is desired
 - EUR commercial FX after 2–3 engagement calibration cycles
 
 ## Pilot evidence

docs/agency-framework.md

@@ -332,9 +332,11 @@ kaizen-agentic schedule prepare coach   # orientation bundle for a scheduled run
 `schedule prepare <agent>` bundles the agent prompt, project memory, metrics
 summary, and repo pointers — offline, no State Hub required. kaizen-agentic does
 **not** run cron, own durable work, or invoke an LLM. activity-core fires the
-schedule and creates `ops_run` work; rein-aharness claims it and executes the
+schedule and creates `ops_run` work. Task intake remains rein-local where
-prepared bundle through a governed `agent-session`. Manual sessions remain a
+implemented. The blueprint/instance supplies a versioned
-supported fallback.
+`harness_profile_ref`; Glas resolves the concrete rein, model route, sandbox,
+tool policy, and limits and returns normalized evidence. Manual sessions remain
+a supported fallback.
 
 Schema: [integrations/schedule-schema.md](integrations/schedule-schema.md).
 

docs/assessments/2026-08-20-intent-role-automation-readiness.md

@@ -4,6 +4,15 @@
 
 **Workplan:** KAIZEN-WP-0012
 
+> **Execution-boundary update, 2026-08-21:** This assessment captured the
+> then-live activity-core → rein-aharness intake path. GLAS-WP-0004 subsequently
+> proved contract 1.0 across two rein/model constellations. Kaizen blueprints
+> and agent instances should now hand a versioned `harness_profile_ref` to
+> glas-harness rather than encode a concrete rein, provider, model, or sandbox.
+> Scheduling remains with activity-core and task/blueprint intake remains
+> caller- or rein-local; the assessment's ownership and budget findings remain
+> valid.
+
 **Assessment:** materially aligned, operationally useful, not yet at the full
 `INTENT.md` maturity target
 

docs/integrations/activity-core-handoff-wp0006.md

@@ -1,15 +1,18 @@
 # activity-core Handoff — Scheduled Agent Execution (WP-0006)
 
 > **Historical supplier handoff, completed for the initial resolver/pilot path.**
-> The current production composition is activity-core durable `ops_run` work
+> The initial production composition used activity-core durable `ops_run` work
-> claimed by rein-aharness through its generic `agent-session` approach. The
+> claimed by rein-aharness through its generic `agent-session` approach. Glas
-> resolver and preparation contracts below remain compatible inputs.
+> contract 1.0 now defines the rein-neutral execution boundary: new blueprint
+> and agent-instance handoffs should carry a versioned `harness_profile_ref`.
+> The resolver and preparation contracts below remain compatible inputs.
 
 Coordination checklist for the **activity-core** team to enable kaizen scheduled
 agent runs. kaizen-agentic owns the schedule contract, the prepare CLI, and the
 ActivityDefinition **drafts**; activity-core owns the resolver, schedule firing,
-and durable work creation (repo boundary, ADR-005); rein-aharness owns governed
+and durable work creation (repo boundary, ADR-005). Task intake remains
-execution.
+rein-local where implemented; glas-harness owns explicit execution-profile
+resolution and normalized evidence.
 
 Open this as an activity-core issue/PR titled *"Enable kaizen scheduled agent
 execution (WP-0006)"* and track the boxes there.
@@ -74,9 +77,9 @@ kaizen-agentic schedule prepare coach   # non-empty orientation bundle
 
 Then in activity-core: run the resolver (dry-run) and confirm one
 `scheduled_run` per enabled `(repo, agent)` with a correct `prepare_command`.
-For unattended operation, confirm the resulting `ops_run` is claimable and the
+For unattended operation, confirm the resulting `ops_run` is claimable, the
-rein-aharness worker selects the `agent-session` approach before enabling the
+blueprint/instance provides an approved `harness_profile_ref`, and Glas resolves
-schedule broadly.
+it without a hidden rein/model fallback before enabling the schedule broadly.
 
 ## Pilot roster
 

docs/integrations/forward-deployed-engagement-playbook.md

@@ -90,7 +90,7 @@ kaizen-agentic engagement close-session eng-... \
 - Enable entries in engagement `schedule.yml`; they declare intent but do not
   activate unattended work by themselves
 - Continue manual prepare until an engagement-specific activity-core definition
-  and rein-aharness profile have passed a bounded smoke test
+  and versioned Glas execution profile have passed a bounded smoke test
 - Daily: health + load; weekly: OS/security pass (privileged changes still gated)
 
 ### 8. Ramp-down (when ending)
@@ -161,9 +161,12 @@ Not required for operate. When ready:
 
 1. Customer or engagement repo owns the ActivityDefinition that identifies the
    engagement and emits idempotent, claimable `ops_run` work in activity-core.
-2. The run carries `agent-session` approach metadata and a compatible
+2. The run or bound agent instance carries organizational references and an
-   rein-aharness profile reference; it does not carry secrets or raw prompts.
+   approved versioned `harness_profile_ref`; it does not carry secrets or raw
-3. rein-aharness claims the run, resolves credentials and policy, and executes
+   provider prompts.
+3. Rein-local intake or another caller claims/receives the work and constructs
+   the Glas `ExecutionRequest`. Glas resolves the concrete rein, model, sandbox,
+   tool policy, and limits; the selected rein executes
    `kaizen-agentic engagement prepare <id>` inside the bounded session.
 4. Completion links the `ops_run`, engagement report, metrics, and State Hub
    progress without duplicating confidential content into coordination records.

engagements/pilots/eng-coulomb-railiance01-ho-001/schedule.yml

@@ -1,6 +1,6 @@
 # Duty cadence for eng-coulomb-railiance01-ho-001
 # Entries declare intent. Manual prepare remains active until an engagement-specific
-# activity-core definition and rein-aharness profile are registered and smoke-tested.
+# activity-core definition and versioned Glas profile are registered and smoke-tested.
 engagement_id: eng-coulomb-railiance01-ho-001
 timezone: Europe/Berlin
 entries:
@@ -13,7 +13,7 @@ entries:
     cadence: daily
     cron: "0 9 * * 1-5"
     enabled: true
-    notes: "Declared after T10 ramp-up exit 2026-07-16; manual prepare until activity-core/rein-aharness wiring is accepted"
+    notes: "Declared after T10 ramp-up exit 2026-07-16; manual prepare until activity-core/Glas wiring is accepted"
 
   - id: weekly-os-security
     agent: host-operator

registry/capabilities/capability.agents.kaizen-framework.md

@@ -45,7 +45,7 @@ external_evidence:
     confidence: low
     basis: consumer_quality_signals
     known_reliability_risks:
-    - downstream execution depends on activity-core and rein-aharness contracts outside this repository
+    - downstream execution depends on activity-core, glas-harness, and selected-rein contracts outside this repository
 discovery:
   intent: Let agents arrive informed, work within explicit roles and engagement boundaries,
     and improve through project memory, execution evidence, and governed scheduling contracts.

workplans/ADHOC-2026-08-21.md

@@ -0,0 +1,27 @@
+---
+id: ADHOC-2026-08-21
+type: workplan
+title: "Ad hoc Glas execution-boundary alignment"
+domain: agents
+repo: kaizen-agentic
+status: finished
+owner: codex
+topic_slug: glas-execution-boundary-alignment
+created: "2026-08-21"
+updated: "2026-08-21"
+---
+
+# Ad hoc Glas execution-boundary alignment
+
+## Align the documented execution handoff with Glas contract 1.0
+
+```task
+id: ADHOC-2026-08-21-T01
+status: done
+priority: high
+```
+
+Apply the completed GLAS-WP-0004 handoff: Kaizen blueprints and agent instances
+name a versioned `harness_profile_ref`; Glas resolves the concrete rein, model,
+sandbox, tool policy, and limits. Preserve activity-core scheduling and durable
+work ownership and rein-local task intake where it already exists.

workplans/KAIZEN-WP-0014-repository-consolidation.md

@@ -59,8 +59,9 @@ status: done
 priority: high
 ```
 
-Update the activity-core integration material to describe the durable `ops_run` and
+Update the activity-core integration material to describe the durable `ops_run`,
-rein-aharness execution path while retaining manual execution as a supported fallback.
+versioned Glas profile, and selected-rein execution path while retaining manual
+execution as a supported fallback.
 
 ## Resolve pilot-vault custody