[AUTO-ARCH] Architecture Review (Cycle 7) #6242

Closed
opened 2026-04-09 18:09:58 +00:00 by HAL9000 · 0 comments
Owner

Architecture Supervisor Status — 2026-04-09 18:10

Agent: architect
Instance: architect-1
Cycle: 7
Status: active

Summary

Continuous architecture monitoring active. Monitoring milestones v3.2.0 through v3.7.0 for spec coverage gaps, implementer ambiguities, and human requests.

Work Completed This Cycle (Cycle 7)

New Issues Assessed

Issue Title Assessment
#6237 agents action create CLI drops invariants, estimation_actor, invariant_actor, automation_profile Implementation gap — spec §33333-33343 clearly documents all four fields in action YAML schema; no spec change needed
#6231 Custom automation profiles fall back to manual in _resolve_profile_for_plan Implementation gap — spec §28594-28600 clearly documents custom profile resolution; no spec change needed
#6226 agents plan use --automation-profile rejects custom profiles Implementation gap — spec §28600 explicitly shows local/careful-auto as valid profile name; no spec change needed

New Spec Proposals Assessed

No new spec proposals since Cycle 6 (last checked: 2026-04-09T17:51:00Z).

Architectural Clarifications Posted

None this cycle — new issues are all implementation gaps with clear spec backing.

All Open Spec PRs (Awaiting Human Approval)

PR Title Status
#4583 Correct SandboxStrategy protocol name, write() return type Open, Needs Feedback
#4805 Add Milestone Plan section to specification Open, Needs Feedback
#4932 Add Milestone Plan section for v3.2.0 through v3.7.0 Open, Needs Feedback
#4979 Clarify TUI session export formats and persona schema fields Open, Needs Feedback
#5035 Architecture corrections cycle 1 — invariant precedence, TUI shell safety, sandbox protocol Open, Needs Feedback
#5133 Architecture corrections cycle 2 — plan revert/resume, correction_attempts DDL Open, Needs Feedback
#5163 Align checkpoint trigger names and config key path Open, Needs Feedback
#5217 Align ARCE acronym, max-rounds default, TUI preset key binding Open, Needs Feedback
#5325 Add Groq, Cohere, Together, Mistral, Ollama provider documentation Open, Needs Feedback
#5393 Clarify agents validation attach synopsis (superseded by #6211) Open, Needs Feedback
#5403 Extend agents diagnostics example to show all 9 supported providers Open, Needs Feedback
#5488 Align subgraph node field from actor_path to actor_ref Open, Needs Feedback
#5595 Clarify agents plan correct accepts plan ID or decision ID Open, Needs Feedback
#5655 Add agents plan revert and agents plan resume commands Open, Needs Feedback
#5658 Align custom resource type YAML format with schema definition Open, Needs Feedback
#5877 Update Azure OpenAI API version default to 2024-05-01-preview Open, Needs Feedback
#5905 Fix invariant precedence chain reference Open, Needs Feedback
#5914 Align AIProviderInterface with implementation Open, Needs Feedback
#5966 Architecture corrections cycle 3 — sandbox state persistence, LangGraph TypedDict Open, Needs Feedback
#6004 Document ReconciliationBlockedError (original) Open, Needs Feedback
#6007 Document ReconciliationBlockedError (duplicate — human should close one) Open, Needs Feedback
#6180 Document context_tier_hydrator module in ACMS architecture section Open, Needs Feedback
#6208 Add agents plan show command specification Open, Needs Feedback
#6209 Document ACMS built-in context strategy retrieval algorithms Open, Needs Feedback
#6210 Document GitWorktreeSandbox as canonical sandbox for git-checkout resources Open, Needs Feedback
#6211 Align agents validation attach extra args to --key value format Open, Needs Feedback

⚠️ Action Required: 26 spec PRs are awaiting human approval. PRs #6004 and #6007 are duplicates — one should be closed. PR #5393 is superseded by #6211.

Milestone Spec Coverage Assessment

Milestone Open Issues Closed Issues Spec Coverage
v3.2.0 278 248 Covered (decisions, invariants, corrections)
v3.3.0 108 108 Covered — PR #6208 adds missing plan show command
v3.4.0 102 132 Covered — PR #6209 adds ACMS strategy algorithms; PR #6210 adds GitWorktreeSandbox
v3.5.0 844 196 Covered (autonomy hardening, A2A facade, guards) — 844 open issues are implementation gaps
v3.6.0 231 150 Covered (advanced concepts, container execution, plugins)
v3.7.0 513 422 Covered (TUI, ADR-044/045/046)

Key Architectural Concerns (Active)

  1. 26 spec PRs awaiting human approval — Implementers working from current master; pending PRs contain important corrections
  2. Custom automation profile resolution broken (#6226, #6231) — Spec §28594-28600 is correct; implementation must use AutomationProfileService.get_profile() instead of BUILTIN_PROFILES dict lookup
  3. agents action create drops 4 fields (#6237) — Spec §33333-33343 is correct; CLI must pass invariants, estimation_actor, invariant_actor, automation_profile to service.create_action()
  4. A2A Facade completeness (Epic #6205) — 7 child issues; spec §A2A is correct; all are implementation gaps
  5. Duplicate ReconciliationBlockedError PRs (#6004, #6007) — Human should close one
  6. Spec restructure proposal (#6195) — 47k line spec is 15× over threshold; deferred pending human decision

Architectural Guidance for Implementers

Custom Automation Profile Resolution (Issues #6226, #6231)

The spec (§28594) is clear: custom profiles must be resolved via AutomationProfileService, not via BUILTIN_PROFILES dict. The fix pattern is:

# WRONG (current implementation):
return BUILTIN_PROFILES.get(profile_name, BUILTIN_PROFILES["manual"])

# CORRECT (per spec §28594):
try:
    return self._automation_profile_service.get_profile(profile_name)
except NotFoundError:
    logger.warning("Unknown automation profile %s, falling back to manual", profile_name)
    return BUILTIN_PROFILES["manual"]

This applies to both _resolve_profile_for_plan() in plan_lifecycle_service.py and the validation in plan.py CLI command.

Action Create Missing Fields (Issue #6237)

The spec (§33333-33343) documents all four fields as part of the action YAML schema. The CLI must pass them to service.create_action(). No spec change needed — this is a straightforward implementation fix.

Next Actions

  • Continue monitoring for new spec proposals and ambiguities
  • Check for new UAT issues indicating spec gaps
  • Monitor for human requests mentioning @architect
  • Next tracking update in ~10 cycles

Automated by CleverAgents Bot
Supervisor: Architecture | Agent: architect | Instance: architect-1

# Architecture Supervisor Status — 2026-04-09 18:10 **Agent**: architect **Instance**: architect-1 **Cycle**: 7 **Status**: active ## Summary Continuous architecture monitoring active. Monitoring milestones v3.2.0 through v3.7.0 for spec coverage gaps, implementer ambiguities, and human requests. ## Work Completed This Cycle (Cycle 7) ### New Issues Assessed | Issue | Title | Assessment | |-------|-------|------------| | #6237 | `agents action create` CLI drops invariants, estimation_actor, invariant_actor, automation_profile | **Implementation gap** — spec §33333-33343 clearly documents all four fields in action YAML schema; no spec change needed | | #6231 | Custom automation profiles fall back to manual in `_resolve_profile_for_plan` | **Implementation gap** — spec §28594-28600 clearly documents custom profile resolution; no spec change needed | | #6226 | `agents plan use --automation-profile` rejects custom profiles | **Implementation gap** — spec §28600 explicitly shows `local/careful-auto` as valid profile name; no spec change needed | ### New Spec Proposals Assessed No new spec proposals since Cycle 6 (last checked: 2026-04-09T17:51:00Z). ### Architectural Clarifications Posted None this cycle — new issues are all implementation gaps with clear spec backing. ## All Open Spec PRs (Awaiting Human Approval) | PR | Title | Status | |----|-------|--------| | #4583 | Correct SandboxStrategy protocol name, write() return type | Open, Needs Feedback | | #4805 | Add Milestone Plan section to specification | Open, Needs Feedback | | #4932 | Add Milestone Plan section for v3.2.0 through v3.7.0 | Open, Needs Feedback | | #4979 | Clarify TUI session export formats and persona schema fields | Open, Needs Feedback | | #5035 | Architecture corrections cycle 1 — invariant precedence, TUI shell safety, sandbox protocol | Open, Needs Feedback | | #5133 | Architecture corrections cycle 2 — plan revert/resume, correction_attempts DDL | Open, Needs Feedback | | #5163 | Align checkpoint trigger names and config key path | Open, Needs Feedback | | #5217 | Align ARCE acronym, max-rounds default, TUI preset key binding | Open, Needs Feedback | | #5325 | Add Groq, Cohere, Together, Mistral, Ollama provider documentation | Open, Needs Feedback | | #5393 | Clarify agents validation attach synopsis (superseded by #6211) | Open, Needs Feedback | | #5403 | Extend agents diagnostics example to show all 9 supported providers | Open, Needs Feedback | | #5488 | Align subgraph node field from actor_path to actor_ref | Open, Needs Feedback | | #5595 | Clarify agents plan correct accepts plan ID or decision ID | Open, Needs Feedback | | #5655 | Add agents plan revert and agents plan resume commands | Open, Needs Feedback | | #5658 | Align custom resource type YAML format with schema definition | Open, Needs Feedback | | #5877 | Update Azure OpenAI API version default to 2024-05-01-preview | Open, Needs Feedback | | #5905 | Fix invariant precedence chain reference | Open, Needs Feedback | | #5914 | Align AIProviderInterface with implementation | Open, Needs Feedback | | #5966 | Architecture corrections cycle 3 — sandbox state persistence, LangGraph TypedDict | Open, Needs Feedback | | #6004 | Document ReconciliationBlockedError (original) | Open, Needs Feedback | | #6007 | Document ReconciliationBlockedError (duplicate — human should close one) | Open, Needs Feedback | | #6180 | Document context_tier_hydrator module in ACMS architecture section | Open, Needs Feedback | | #6208 | Add `agents plan show` command specification | Open, Needs Feedback | | #6209 | Document ACMS built-in context strategy retrieval algorithms | Open, Needs Feedback | | #6210 | Document GitWorktreeSandbox as canonical sandbox for git-checkout resources | Open, Needs Feedback | | #6211 | Align `agents validation attach` extra args to --key value format | Open, Needs Feedback | **⚠️ Action Required**: 26 spec PRs are awaiting human approval. PRs #6004 and #6007 are duplicates — one should be closed. PR #5393 is superseded by #6211. ## Milestone Spec Coverage Assessment | Milestone | Open Issues | Closed Issues | Spec Coverage | |-----------|-------------|---------------|---------------| | v3.2.0 | 278 | 248 | ✅ Covered (decisions, invariants, corrections) | | v3.3.0 | 108 | 108 | ✅ Covered — PR #6208 adds missing `plan show` command | | v3.4.0 | 102 | 132 | ✅ Covered — PR #6209 adds ACMS strategy algorithms; PR #6210 adds GitWorktreeSandbox | | v3.5.0 | 844 | 196 | ✅ Covered (autonomy hardening, A2A facade, guards) — 844 open issues are implementation gaps | | v3.6.0 | 231 | 150 | ✅ Covered (advanced concepts, container execution, plugins) | | v3.7.0 | 513 | 422 | ✅ Covered (TUI, ADR-044/045/046) | ## Key Architectural Concerns (Active) 1. **26 spec PRs awaiting human approval** — Implementers working from current master; pending PRs contain important corrections 2. **Custom automation profile resolution broken** (#6226, #6231) — Spec §28594-28600 is correct; implementation must use `AutomationProfileService.get_profile()` instead of `BUILTIN_PROFILES` dict lookup 3. **`agents action create` drops 4 fields** (#6237) — Spec §33333-33343 is correct; CLI must pass `invariants`, `estimation_actor`, `invariant_actor`, `automation_profile` to `service.create_action()` 4. **A2A Facade completeness** (Epic #6205) — 7 child issues; spec §A2A is correct; all are implementation gaps 5. **Duplicate ReconciliationBlockedError PRs** (#6004, #6007) — Human should close one 6. **Spec restructure proposal** (#6195) — 47k line spec is 15× over threshold; deferred pending human decision ## Architectural Guidance for Implementers ### Custom Automation Profile Resolution (Issues #6226, #6231) The spec (§28594) is clear: custom profiles must be resolved via `AutomationProfileService`, not via `BUILTIN_PROFILES` dict. The fix pattern is: ```python # WRONG (current implementation): return BUILTIN_PROFILES.get(profile_name, BUILTIN_PROFILES["manual"]) # CORRECT (per spec §28594): try: return self._automation_profile_service.get_profile(profile_name) except NotFoundError: logger.warning("Unknown automation profile %s, falling back to manual", profile_name) return BUILTIN_PROFILES["manual"] ``` This applies to both `_resolve_profile_for_plan()` in `plan_lifecycle_service.py` and the validation in `plan.py` CLI command. ### Action Create Missing Fields (Issue #6237) The spec (§33333-33343) documents all four fields as part of the action YAML schema. The CLI must pass them to `service.create_action()`. No spec change needed — this is a straightforward implementation fix. ## Next Actions - Continue monitoring for new spec proposals and ambiguities - Check for new UAT issues indicating spec gaps - Monitor for human requests mentioning @architect - Next tracking update in ~10 cycles --- **Automated by CleverAgents Bot** Supervisor: Architecture | Agent: architect | Instance: architect-1
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
cleveragents/cleveragents-core#6242
No description provided.