Quazmoz · Quazmoz · Aug 25, 2026 · Aug 25, 2026 · Aug 25, 2026 · Aug 25, 2026
diff --git a/AGENTS.md b/AGENTS.md
@@ -28,6 +28,8 @@ For `Quazmoz/K8SHomelab`, the specialist must re-read that target repository's c
 
 For codebase maintenance work, the specialist must fingerprint the target repository's actual language/framework/build/test/static-analysis toolchain before editing. It preserves behavior and external contracts by default, requires evidence for risky dead-code/dependency removal, reconciles stale comments in touched code, and performs a second-pass review for fresh agent-generated slop.
 
+For implementation or qualification work that explicitly needs a bounded Integration Owner + independent reviewer loop, use `docs/quickstarts/bounded-completion.md` and `skills/bounded-completion-orchestration.md` after selecting the smallest correct domain owner. Bounded completion is an orchestration overlay, not a replacement owning agent; it cannot widen the selected owner's authority. Distinct-model review counts only when the operator or runtime confirms a genuinely distinct reviewer model.
+
 For Palmier Pro video-editing work, route directly without treating the editing task as generic AI engineering:
 
 | Primary task | Owning agent | Default skill |
@@ -61,8 +63,9 @@ Rules:
 6. Security-focused DevOps routing does not authorize credential, IAM, state, network, controller, or production mutation without explicit task authority.
 7. Documentation authority does not imply permission to change the infrastructure or automation being documented.
 8. Codebase-maintenance authority does not imply permission to change product semantics, external contracts, production data, deployments, or security controls; those require explicit task authority and the appropriate owning specialist when primary.
-9. The K8SHomelab specialist does not infer live-cluster mutation authority from GitHub write access; watched-branch writes can themselves be deployment actions under Flux.
-10. Palmier external MCP tool availability does not imply Palmier in-app agent capabilities are exposed to Claude/Codex.
+9. Bounded-completion orchestration does not create new mutation authority, approvals, or tool permissions; the selected owner and active task remain authoritative.
+10. The K8SHomelab specialist does not infer live-cluster mutation authority from GitHub write access; watched-branch writes can themselves be deployment actions under Flux.
+11. Palmier external MCP tool availability does not imply Palmier in-app agent capabilities are exposed to Claude/Codex.
 
 ## Instruction and Authority Precedence
 
@@ -110,6 +113,7 @@ Update canonical behavior at its canonical source. A wrapper may summarize, rout
 - `README.md` - project overview and usage.
 - `docs/quickstarts/codex.md` - OpenAI Codex usage.
 - `docs/quickstarts/claude.md` - Claude Code usage.
+- `docs/quickstarts/bounded-completion.md` - bounded Integration Owner/reviewer orchestration, durable evidence, and completion-gate usage.
 - `docs/quickstarts/kubernetes-homelab-engineer.md` - repository-specific K8SHomelab usage.
 - `docs/quickstarts/devsecops-security-engineer.md` - DevSecOps security usage for Terraform/OpenTofu, Ansible/AAP, Jenkins, CI/CD, IAM, and supply chain.
 - `docs/quickstarts/devops-documentation-engineer.md` - DevOps documentation-as-code usage.
@@ -122,6 +126,7 @@ Update canonical behavior at its canonical source. A wrapper may summarize, rout
 - `scripts/validate-documentation-stack.py` - DevOps documentation stack contract validation.
 - `scripts/validate-devsecops-security-stack.py` - DevSecOps security stack contract validation.
 - `scripts/validate-codebase-maintenance-stack.py` - codebase-maintenance/de-slop stack contract and routing validation.
+- `scripts/validate-bounded-completion.py` - bounded-completion control-plane and adversarial regression validation.
 
 ## Working Rules
 
@@ -144,7 +149,13 @@ After AgentDefaults changes, run the canonical validation suite:
 python3 scripts/validate-agentdefaults.py
 ```
 
-Its component validators include cross-tool routing, principal engineering contracts, specialist documentation-stack validation, DevSecOps security-stack validation, and codebase-maintenance/de-slop stack validation. Run additional domain-specific tests when the change affects canonical agents, skills, schemas, examples, or executable behavior.
+Its component validators include cross-tool routing, principal engineering contracts, specialist documentation-stack validation, DevSecOps security-stack validation, codebase-maintenance/de-slop validation, and bounded-completion control-plane regressions. Run additional domain-specific tests when the change affects canonical agents, skills, schemas, examples, or executable behavior.
+
+For bounded-completion changes, also review the behavioral cases in:
+
+```text
+docs/bounded-completion-acceptance-tests.md
+```
 
 For K8SHomelab agent changes, also review the behavioral cases in:
 

diff --git a/ENGINEERING_AGENTS_INDEX.md b/ENGINEERING_AGENTS_INDEX.md
@@ -33,6 +33,20 @@ Do not use this index as a substitute for the selected canonical agent. Tool wra
 | Design, build, or audit another reusable agent | [`agents/agent-architect-builder.md`](agents/agent-architect-builder.md) | `skills/agent-design-and-build.md` |
 | Select which automation platform/product should own a workload | [`agents/automation-platform-selection-advisor.md`](agents/automation-platform-selection-advisor.md) | Load only task-relevant platform-selection skills |
 
+## Optional Bounded Completion Orchestration
+
+After selecting the smallest correct owner, implementation or qualification work may add the bounded completion overlay when it needs a single Integration Owner, independent adversarial review, durable evidence, bounded retries/continuation, and an objective completion gate.
+
+```text
+docs/quickstarts/bounded-completion.md
+agents/bounded-completion-lead.md
+agents/bounded-completion-reviewer.md
+skills/bounded-completion-orchestration.md
+scripts/bounded-completion.py
+```
+
+This overlay does **not** become the domain owner and does not widen authority, approvals, or tool permissions. The selected engineering owner remains responsible for domain behavior and safety boundaries. The lead coordinates implementation and evidence; the reviewer challenges plans, findings, visual artifacts when applicable, and final completion claims. Distinct-model evidence is valid only when the operator or runtime confirms that the reviewer actually ran on a different model.
+
 ### Boundary examples
 
 - Work whose target repository is `Quazmoz/K8SHomelab` and whose primary concern is its Kubernetes/Flux/runtime platform -> Kubernetes Homelab Engineer.
@@ -194,7 +208,8 @@ Examples:
 9. Security-focused DevOps tasks may inspect broad platform evidence without turning the specialist into the default owner for non-security refactoring.
 10. Documentation tasks may load platform implementation evidence without inheriting platform mutation authority.
 11. Codebase-maintenance tasks may inspect broad source/tooling evidence but must hand off primary domain/security/platform defects when the required remediation exceeds behavior-preserving maintenance scope.
-12. Task evidence outranks generic guidance; current official documentation outranks stale platform assumptions.
+12. Add bounded completion only after owner selection, and treat it as orchestration/evidence control rather than a new ownership route.
+13. Task evidence outranks generic guidance; current official documentation outranks stale platform assumptions.
 
 ## Shared Invariants
 

diff --git a/README.md b/README.md
@@ -42,6 +42,7 @@ Then choose a path:
 |---|---|
 | Pick the right files quickly | [`docs/user-guide.md`](docs/user-guide.md) |
 | Design, build, or audit another AI agent | [`docs/quickstarts/agent-builder.md`](docs/quickstarts/agent-builder.md) |
+| Run implementation/qualification through a bounded lead + reviewer loop | [`docs/quickstarts/bounded-completion.md`](docs/quickstarts/bounded-completion.md) |
 | De-slop/refactor an existing codebase without silently changing behavior | [`docs/quickstarts/codebase-maintenance-engineer.md`](docs/quickstarts/codebase-maintenance-engineer.md) |
 | Select or challenge an automation platform architecture | [`AUTOMATION_PLATFORM_INDEX.md`](AUTOMATION_PLATFORM_INDEX.md) |
 | Use a local repo-aware coding CLI | [`docs/quickstarts/cli.md`](docs/quickstarts/cli.md) |
@@ -98,6 +99,28 @@ docs/patterns/agent.md
 
 The stack defaults to one agent plus selectively loaded skills. It introduces multiple agents only when permission isolation, independent specialist context, parallel execution with reconciliation, independent verification, separate durable control loops, or fault isolation provides a concrete benefit. A skill, retrieved document, tool output, or sub-agent cannot broaden the parent agent's authority.
 
+### Bounded Completion Stack
+
+Use when implementation or qualification needs a single Integration Owner, independent adversarial review, durable task/findings state, deterministic verification evidence, bounded continuation, and an objective completion gate.
+
+```text
+docs/quickstarts/bounded-completion.md
+agents/bounded-completion-lead.md
+agents/bounded-completion-reviewer.md
+skills/bounded-completion-orchestration.md
+schemas/bounded-completion-task.schema.json
+schemas/bounded-completion-state.schema.json
+schemas/bounded-completion-findings.schema.json
+examples/bounded-completion-task.json
+config/bounded-completion.json
+scripts/bounded-completion.py
+scripts/validate-bounded-completion.py
+.github/agents/bounded-completion-lead.agent.md
+.github/agents/bounded-completion-reviewer.agent.md
+```
+
+Bounded completion is an **orchestration overlay, not a domain owner**. Select the smallest correct engineering owner first; the lead coordinates that owner’s implementation behavior while the reviewer challenges plans, findings, visual evidence when required, and final completion claims. The committed Copilot agents intentionally omit guessed `model:` identifiers, and distinct-model review counts only when the operator or runtime confirms that a genuinely different reviewer model ran.
+
 ### Codebase Maintenance and De-Slop Stack
 
 Use after rapid or agentic development when a repository works but is accumulating quiet maintenance rot: stale comments/docstrings/TODOs, session-to-session duplicate helpers, abandoned implementation paths, abstraction inflation, catch-all failure handling, brittle tests, dependency/configuration residue, or practical performance problems.
@@ -310,7 +333,7 @@ docs/benchmarks/token-efficiency-fresh-2026-06-25.md
 | Claude / Claude Code | [`CLAUDE.md`](CLAUDE.md) | Claude-oriented wrapper, delegates shared rules to `AGENTS.md`. |
 | Gemini / Gemini CLI | [`GEMINI.md`](GEMINI.md) | Gemini-oriented wrapper, delegates shared behavior to `AGENTS.md`. |
 | GitHub Copilot repo instructions | [`.github/copilot-instructions.md`](.github/copilot-instructions.md) | Repo-wide behavior and maintenance rules. |
-| GitHub Copilot custom agents | [`.github/agents/`](.github/agents/) | Selectable profile wrappers. |
+| GitHub Copilot custom agents | [`.github/agents/`](.github/agents/) | Selectable profile wrappers, including bounded completion lead/reviewer roles. |
 | Cursor | [`.cursor/rules/agentdefaults.mdc`](.cursor/rules/agentdefaults.mdc) | Thin rule wrapper pointing back to canonical files. |
 | Windsurf | [`.windsurfrules`](.windsurfrules) | Thin wrapper pointing back to canonical files. |
 | MCP-connected apps | [`docs/quickstarts/palmierpro-mcp.md`](docs/quickstarts/palmierpro-mcp.md) | Palmier Pro video-editing workflow over local MCP. |
@@ -322,6 +345,8 @@ docs/benchmarks/token-efficiency-fresh-2026-06-25.md
 | Type | Name | Path |
 |---|---|---|
 | Agent | Agent Architect and Builder | [`agents/agent-architect-builder.md`](agents/agent-architect-builder.md) |
+| Agent | Bounded Completion Lead | [`agents/bounded-completion-lead.md`](agents/bounded-completion-lead.md) |
+| Agent | Bounded Completion Reviewer | [`agents/bounded-completion-reviewer.md`](agents/bounded-completion-reviewer.md) |
 | Agent | Codebase Maintenance and De-Slop Engineer | [`agents/codebase-maintenance-engineer.md`](agents/codebase-maintenance-engineer.md) |
 | Agent | Automation Platform Selection Advisor | [`agents/automation-platform-selection-advisor.md`](agents/automation-platform-selection-advisor.md) |
 | Agent | Palmier Pro MCP Video Editor | [`agents/palmierpro-mcp-video-editor-agent.md`](agents/palmierpro-mcp-video-editor-agent.md) |
@@ -338,6 +363,7 @@ docs/benchmarks/token-efficiency-fresh-2026-06-25.md
 | Agent | Android Wear OS Release Engineer | [`agents/android-wearos-release-engineer.md`](agents/android-wearos-release-engineer.md) |
 | Agent | US to Europe Travel Advisor | [`agents/us-europe-travel-advisor.md`](agents/us-europe-travel-advisor.md) |
 | Skill | Agent Design and Build | [`skills/agent-design-and-build.md`](skills/agent-design-and-build.md) |
+| Skill | Bounded Completion Orchestration | [`skills/bounded-completion-orchestration.md`](skills/bounded-completion-orchestration.md) |
 | Skill | Codebase De-Slop and Refactoring | [`skills/codebase-de-slop-and-refactoring.md`](skills/codebase-de-slop-and-refactoring.md) |
 | Skill | Automation Platform Capability Taxonomy | [`skills/automation-platform-capability-taxonomy.md`](skills/automation-platform-capability-taxonomy.md) |
 | Skill | Automation Platform Candidate Discovery | [`skills/automation-platform-candidate-discovery.md`](skills/automation-platform-candidate-discovery.md) |
@@ -400,6 +426,7 @@ For the full list, use [`INDEX.md`](INDEX.md).
 | Recipe | Use |
 |---|---|
 | [`examples/agent-build-brief.yaml`](examples/agent-build-brief.yaml) | Ready-to-adapt structured brief for building a repository maintenance agent with explicit runtime and authority boundaries. |
+| [`examples/bounded-completion-task.json`](examples/bounded-completion-task.json) | Structured bounded-completion contract with explicit criteria, verification, approvals, and iteration limits. |
 | [`examples/codebase-maintenance-task.yaml`](examples/codebase-maintenance-task.yaml) | Behavior-preserving de-slop brief covering comments, duplication, dead code, failure semantics, tests, dependencies, and verification. |
 | [`examples/automation-platform-decision-brief.yaml`](examples/automation-platform-decision-brief.yaml) | Category-aware automation selection brief with incumbents, alternatives, hosting, licensing, migration, and evidence controls. |
 | [`examples/app-market-research-brief.yaml`](examples/app-market-research-brief.yaml) | Ready-to-adapt Wear OS market-research brief. |
@@ -439,6 +466,7 @@ It checks:
 - The Agent Architect and Builder stack preserves canonical build modes, architecture choices, permission classes, runtime-capability fields, required contract terms, and all 22 acceptance scenarios.
 - The automation platform stack preserves its canonical capability, evidence, economics, and acceptance-test invariants.
 - The Codebase Maintenance and De-Slop stack preserves behavior-change authorization, comment reconciliation, mutation/verification contracts, performance-evidence honesty, routing, and adversarial acceptance coverage.
+- The bounded-completion stack preserves deterministic task/findings state, verification freshness, review/integrity evidence, approval provenance, visual-artifact requirements, iteration/stop bounds, and completion-gate regressions.
 - Local Markdown links resolve.
 
 ## Benchmark Evidence
@@ -460,15 +488,21 @@ agentdefaults/
 ├── AUTOMATION_PLATFORM_INDEX.md
 ├── agentdefaults.manifest.json
 ├── scripts/
+│   ├── bounded-completion.py
 │   ├── validate-agentdefaults.py
+│   ├── validate-bounded-completion.py
 │   └── validate-codebase-maintenance-stack.py
 ├── schemas/
 │   ├── agent-build-brief.schema.json
+│   ├── bounded-completion-task.schema.json
+│   ├── bounded-completion-state.schema.json
+│   ├── bounded-completion-findings.schema.json
 │   ├── codebase-maintenance-task.schema.json
 │   ├── automation-platform-decision-brief.schema.json
 │   └── app-market-research-brief.schema.json
 ├── docs/
 │   ├── agent-builder-acceptance-tests.md
+│   ├── bounded-completion-acceptance-tests.md
 │   ├── codebase-maintenance-engineer-acceptance-tests.md
 │   ├── automation-platform-selection-acceptance-tests.md
 │   ├── user-guide.md
@@ -477,6 +511,7 @@ agentdefaults/
 │   ├── app-market-research-acceptance-tests.md
 │   ├── quickstarts/
 │   │   ├── agent-builder.md
+│   │   ├── bounded-completion.md
 │   │   ├── codebase-maintenance-engineer.md
 │   │   ├── automation-platform-selection.md
 │   │   └── community-app-validation.md
@@ -485,15 +520,19 @@ agentdefaults/
 │   └── benchmarks/
 ├── examples/
 │   ├── agent-build-brief.yaml
+│   ├── bounded-completion-task.json
 │   ├── codebase-maintenance-task.yaml
 │   └── automation-platform-decision-brief.yaml
 ├── agents/
 │   ├── agent-architect-builder.md
+│   ├── bounded-completion-lead.md
+│   ├── bounded-completion-reviewer.md
 │   ├── codebase-maintenance-engineer.md
 │   ├── automation-platform-selection-advisor.md
 │   └── community-app-idea-validation-agent.md
 ├── skills/
 │   ├── agent-design-and-build.md
+│   ├── bounded-completion-orchestration.md
 │   ├── codebase-de-slop-and-refactoring.md
 │   ├── automation-platform-capability-taxonomy.md
 │   ├── automation-platform-candidate-discovery.md
@@ -503,18 +542,24 @@ agentdefaults/
 │   ├── gitops-runbook-and-workflow-platform-analysis.md
 │   └── subreddit-app-idea-validation.md
 ├── prompts/
+│   ├── orchestration/
+│   │   ├── start-bounded-completion.md
+│   │   └── resume-bounded-completion.md
 │   ├── implementation/
 │   │   └── codebase-de-slop-task.md
 │   ├── planning/
 │   │   ├── build-ai-agent.md
 │   │   └── select-automation-platform.md
 │   ├── review/
+│   │   ├── bounded-completion-review.md
 │   │   └── challenge-automation-platform-choice.md
 │   └── research/
 │       └── validate-app-idea-in-community.md
 ├── .github/
 │   └── agents/
 │       ├── agent-architect-builder.agent.md
+│       ├── bounded-completion-lead.agent.md
+│       ├── bounded-completion-reviewer.agent.md
 │       ├── codebase-maintenance-engineer.agent.md
 │       ├── automation-platform-selection-advisor.agent.md
 │       └── community-app-idea-validator.agent.md
@@ -550,7 +595,7 @@ Use [`docs/patterns/agent.md`](docs/patterns/agent.md) for new agent profiles an
 
 ## Status
 
-Usable cross-tool scaffold with reusable agent-construction defaults, behavior-preserving codebase maintenance/de-slop engineering, category-aware automation platform architecture and selection, token-efficiency defaults, browser-based app-market research, focused subreddit/community app-idea validation, Palmier Pro MCP video-editing defaults, Google Play growth tooling, Wear OS development and release stacks, travel research, tool wrappers, quickstarts, examples, schemas, patterns, and validation tooling.
+Usable cross-tool scaffold with reusable agent-construction defaults, bounded lead/reviewer completion orchestration, behavior-preserving codebase maintenance/de-slop engineering, category-aware automation platform architecture and selection, token-efficiency defaults, browser-based app-market research, focused subreddit/community app-idea validation, Palmier Pro MCP video-editing defaults, Google Play growth tooling, Wear OS development and release stacks, travel research, tool wrappers, quickstarts, examples, schemas, patterns, and validation tooling.
 
 ## License