feat(actor): add plan_subplan tool and decision emission #463

Merged

aditya merged 11 commits from feature/m5-subplan-actor into master 2026-03-05 07:12:22 +00:00

Conversation 5 Commits 11 Files Changed 11

+1570 -4

aditya commented 2026-02-27 08:12:01 +00:00

Member

Copy Link

Copy Source

Description

Implements the builtin/plan-subplan tool that strategy actors call to emit
SUBPLAN_SPAWN or SUBPLAN_PARALLEL_SPAWN decisions when decomposing a large
plan into coordinated child plans. Validates the subplan payload, applies
sensible defaults, generates human-readable rationale text, and optionally
persists the emitted Decision via an injected DecisionService.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Refactoring (no functional changes)
• Documentation update
• Test improvements
• CI/CD changes

Quality Checklist

• Code follows the project's coding standards (see CONTRIBUTING.md)
• All public/protected methods have argument validation
• Static typing is complete (no Any unless justified)
• nox -s typecheck passes with no errors
• nox -s lint passes with no errors
• Unit tests written/updated (Behave scenarios in features/)
• Integration tests written/updated (Robot suites in robot/) if applicable
• Coverage remains above 97% (nox -s coverage_report)
• No security issues introduced (nox -s security_scan)
• No dead code introduced (nox -s dead_code)
• Documentation updated if behavior changed

Testing

All tests written before implementation (TDD). Tests cover validation,
defaults, serial vs parallel decision types, rationale generation, decision
service injection, registry registration, and ToolRunner execution.

Test Commands Run

nox -s unit_tests       # 20 Behave scenarios, 70 steps — all passed
nox -s typecheck        # 0 errors, 0 warnings
nox -s lint             # 0 errors (ruff, bandit)
nox -s coverage_report  # 100% coverage on subplan_tool.py
python -m robot robot/plan_subplan_tool.robot  # 9 smoke tests — all passed

Related Issues

ISSUES CLOSED #198

Implementation Notes

New files

File	Purpose
src/cleveragents/tool/builtins/subplan_tool.py	Core tool — SubplanPayload, make_plan_subplan_spec, PLAN_SUBPLAN_SPEC
features/plan_subplan_tool.feature	Behave BDD scenarios (20 scenarios)
features/steps/plan_subplan_tool_steps.py	Step definitions with _StubDecisionService
robot/plan_subplan_tool.robot	Robot Framework smoke suite (9 tests)
robot/helper_plan_subplan_tool.py	Robot helper script
benchmarks/subplan_actor_tool_bench.py	ASV benchmarks (5 suites)
examples/actors/strategy_with_subplan.yaml	Actor YAML example with serial + parallel payload examples

Modified files

• src/cleveragents/tool/builtins/__init__.py — added register_subplan_tool, SubplanPayload, make_plan_subplan_spec, ALL_SUBPLAN_TOOLS exports
• CHANGELOG.md — added #198 entry

Architecture decisions

Optional DecisionService injection via factory: The tool itself is a
pure function. make_plan_subplan_spec(decision_service=None) creates a
ToolSpec whose handler is a closure bound to the service. In production the
runtime injects the real DecisionService; in tests a list-based stub is used,
keeping 100% of tool logic testable without a database.

SubplanPayload as the single validation layer: All input constraints
(non-empty goal, at least one of resource_scopes/project_ref, max_parallel
1–50, valid merge_strategy enum, valid context_view value) are enforced by a
frozen Pydantic model. Any validation failure returns {"success": False, "error": ...} rather than raising, so the ToolRunner never sees an unhandled exception.

Decision type driven by parallel flag: parallel=False emits
SUBPLAN_SPAWN; parallel=True emits SUBPLAN_PARALLEL_SPAWN. The
dependencies list (sibling subplan ULIDs) is only meaningful for parallel
spawns but is accepted and stored in both cases for forward compatibility.

Rationale auto-generation: _build_rationale() constructs a multi-line
human-readable string from the payload (goal, targeted resources, execution
mode, dependencies, context view override). This text is stored on the
Decision.rationale field and surfaced in plan explain output.

## Description Implements the `builtin/plan-subplan` tool that strategy actors call to emit `SUBPLAN_SPAWN` or `SUBPLAN_PARALLEL_SPAWN` decisions when decomposing a large plan into coordinated child plans. Validates the subplan payload, applies sensible defaults, generates human-readable rationale text, and optionally persists the emitted `Decision` via an injected `DecisionService`. ## Type of Change - [ ] Bug fix (non-breaking change which fixes an issue) - [x] New feature (non-breaking change which adds functionality) - [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected) - [ ] Refactoring (no functional changes) - [x] Documentation update - [ ] Test improvements - [ ] CI/CD changes ## Quality Checklist - [x] Code follows the project's coding standards (see CONTRIBUTING.md) - [x] All public/protected methods have argument validation - [x] Static typing is complete (no `Any` unless justified) - [x] `nox -s typecheck` passes with no errors - [x] `nox -s lint` passes with no errors - [x] Unit tests written/updated (Behave scenarios in `features/`) - [x] Integration tests written/updated (Robot suites in `robot/`) if applicable - [x] Coverage remains above 97% (`nox -s coverage_report`) - [x] No security issues introduced (`nox -s security_scan`) - [x] No dead code introduced (`nox -s dead_code`) - [x] Documentation updated if behavior changed ## Testing All tests written before implementation (TDD). Tests cover validation, defaults, serial vs parallel decision types, rationale generation, decision service injection, registry registration, and ToolRunner execution. ### Test Commands Run ```bash nox -s unit_tests # 20 Behave scenarios, 70 steps — all passed nox -s typecheck # 0 errors, 0 warnings nox -s lint # 0 errors (ruff, bandit) nox -s coverage_report # 100% coverage on subplan_tool.py python -m robot robot/plan_subplan_tool.robot # 9 smoke tests — all passed ``` ## Related Issues ISSUES CLOSED #198 ## Implementation Notes ### New files | File | Purpose | |---|---| | `src/cleveragents/tool/builtins/subplan_tool.py` | Core tool — `SubplanPayload`, `make_plan_subplan_spec`, `PLAN_SUBPLAN_SPEC` | | `features/plan_subplan_tool.feature` | Behave BDD scenarios (20 scenarios) | | `features/steps/plan_subplan_tool_steps.py` | Step definitions with `_StubDecisionService` | | `robot/plan_subplan_tool.robot` | Robot Framework smoke suite (9 tests) | | `robot/helper_plan_subplan_tool.py` | Robot helper script | | `benchmarks/subplan_actor_tool_bench.py` | ASV benchmarks (5 suites) | | `examples/actors/strategy_with_subplan.yaml` | Actor YAML example with serial + parallel payload examples | ### Modified files - `src/cleveragents/tool/builtins/__init__.py` — added `register_subplan_tool`, `SubplanPayload`, `make_plan_subplan_spec`, `ALL_SUBPLAN_TOOLS` exports - `CHANGELOG.md` — added `#198` entry ### Architecture decisions **Optional `DecisionService` injection via factory:** The tool itself is a pure function. `make_plan_subplan_spec(decision_service=None)` creates a `ToolSpec` whose handler is a closure bound to the service. In production the runtime injects the real `DecisionService`; in tests a list-based stub is used, keeping 100% of tool logic testable without a database. **`SubplanPayload` as the single validation layer:** All input constraints (non-empty goal, at least one of `resource_scopes`/`project_ref`, `max_parallel` 1–50, valid `merge_strategy` enum, valid `context_view` value) are enforced by a frozen Pydantic model. Any validation failure returns `{"success": False, "error": ...}` rather than raising, so the ToolRunner never sees an unhandled exception. **Decision type driven by `parallel` flag:** `parallel=False` emits `SUBPLAN_SPAWN`; `parallel=True` emits `SUBPLAN_PARALLEL_SPAWN`. The `dependencies` list (sibling subplan ULIDs) is only meaningful for parallel spawns but is accepted and stored in both cases for forward compatibility. **Rationale auto-generation:** `_build_rationale()` constructs a multi-line human-readable string from the payload (goal, targeted resources, execution mode, dependencies, context view override). This text is stored on the `Decision.rationale` field and surfaced in `plan explain` output.

aditya added this to the v3.4.0 milestone 2026-02-27 08:12:01 +00:00

aditya added the

Type

Feature

label 2026-02-27 08:12:01 +00:00

aditya added 1 commit 2026-02-27 08:12:02 +00:00

feat(actor): add plan_subplan tool and decision emission ... b888afab71

Add builtin/plan-subplan tool for strategy actors to emit SUBPLAN_SPAWN
or SUBPLAN_PARALLEL_SPAWN decisions when decomposing a plan into child
plans. Implements all acceptance criteria from issue #198:

- SubplanPayload (Pydantic) validates goal, resource_scopes/project_ref
  (at least one required), merge_strategy, max_parallel (1-50), parallel
  flag, dependencies, and context_view override.
- Defaults: merge_strategy=git_three_way, max_parallel=5, parallel=False,
  dependencies=[]. Omitted fields inherit sensible values automatically.
- make_plan_subplan_spec(decision_service=None) factory supports optional
  DecisionService injection for persistent decision recording.
- _build_rationale() generates human-readable rationale text (goal, scope,
  execution mode, dependencies, context_view) surfaced in plan explain.
- register_subplan_tool() added to tool/builtins/__init__.py for bulk
  registration. PLAN_SUBPLAN_SPEC exported as the default ready-to-use spec.
- Actor YAML example (examples/actors/strategy_with_subplan.yaml) with
  annotated serial and parallel spawn payload examples.
- Behave BDD: 20 scenarios, 70 steps covering validation, defaults,
  decision type, rationale, service injection, registry, and ToolRunner.
- Robot Framework: 9 smoke tests via robot/plan_subplan_tool.robot.
- ASV benchmarks: benchmarks/subplan_actor_tool_bench.py (5 suites).
- Coverage: 100% on subplan_tool.py. Lint, typecheck, and security clean.

ISSUES CLOSED: #198

aditya requested review from CoreRasurae 2026-02-27 08:12:13 +00:00

aditya added 1 commit 2026-02-27 08:46:33 +00:00

Merge branch 'master' into feature/m5-subplan-actor 2b5b73c689

aditya added 1 commit 2026-02-27 09:28:24 +00:00

fix(tests): update actor_examples count to 7 after subplan example ... 131be21e80

Refs: #198

aditya force-pushed feature/m5-subplan-actor from 131be21e80 to 4c4bd74d94 2026-02-27 09:36:55 +00:00 Compare

aditya force-pushed feature/m5-subplan-actor from 4c4bd74d94 to 131be21e80 2026-02-27 09:38:36 +00:00 Compare

CoreRasurae commented 2026-02-27 11:31:07 +00:00

Member

Copy Link

Copy Source

Code Review — Bugs & Security Findings

Reviewed commits b888afab (feat) and 131be21e (fix) on feature/m5-subplan-actor against issue #198 acceptance criteria and docs/specification.md.

---

BUGS

B1 [High] parent_decision_id accepted but never returned to caller

File: src/cleveragents/tool/builtins/subplan_tool.py:268-281

The parent_decision_id field is accepted in SubplanPayload (line 149), stored in the Decision object (line 255), but the handler's return dict (lines 268-281) omits it entirely. The calling actor has no way to confirm the parent linkage was recorded.

The corresponding Behave scenario ("parent_decision_id is included in the decision when provided" at features/plan_subplan_tool.feature:112-116) only asserts success=True and decision_type — it never verifies parent_decision_id is present in the result, making the test name misleading and the assertion vacuous.

Fix: Add "parent_decision_id": payload.parent_decision_id to the return dict at line 281, and update the test to assert its value.

---

B2 [Medium] "Verify all examples are documented" scenario not updated

File: features/actor_examples.feature:665-677

The fix commit (131be21e) correctly updated the "Verify all example files exist" scenario (line 654) to expect 7 files including strategy_with_subplan.yaml. However, the adjacent "Verify all examples are documented" scenario (line 665) was not updated to check that strategy_with_subplan.yaml is referenced in docs/reference/actors_examples.md. Additionally, docs/reference/actors_examples.md does not appear to have been updated to reference the new example.

---

B3 [Medium] Specification naming discrepancy

Spec: docs/specification.md §Child Plan Spawning Mechanism:

"The local/create-subplan tool is independently registered and referenced by name in the actor graph node."

Implementation: builtin/plan-subplan (subplan_tool.py:76)

These differ in both namespace (local/ vs builtin/) and name (create-subplan vs plan-subplan). Either the specification or the implementation should be updated for consistency.

---

SECURITY

S1 [High] No ULID format validation on plan_id, dependencies, parent_decision_id

File: src/cleveragents/tool/builtins/subplan_tool.py:115,141-143,149-151

SubplanPayload accepts arbitrary strings for plan_id (only min_length=1), dependencies (no validation), and parent_decision_id (no validation). These fields semantically represent ULIDs.

The Decision model in the same codebase (domain/models/core/decision.py) rigorously validates ULID fields with a regex pattern (^[0-9A-HJKMNP-TV-Z]{26}$) and dedicated @field_validator methods. This inconsistency means:

• Malformed plan_id values pass the tool handler, then fail inside Decision(**decision_kwargs), producing a confusing error message about the Decision model rather than about the tool input.
• Malformed dependencies entries flow into the rationale text completely unchecked.

Fix: Add @field_validator methods on plan_id, dependencies, and parent_decision_id in SubplanPayload enforcing the same ULID regex pattern used by the Decision model.

---

S2 [Medium] No path validation on resource_scopes

File: src/cleveragents/tool/builtins/subplan_tool.py:119-121

resource_scopes accepts arbitrary strings with no format or sanitization checks. Values like [""], ["../../etc/passwd"], or extremely long strings are accepted and flow into the Decision rationale and downstream processing. Early validation (non-empty strings, no path traversal sequences) would prevent garbage from entering the decision tree.

---

S3 [Low] Overly broad exception handling masks programming errors

File: src/cleveragents/tool/builtins/subplan_tool.py:226

try:
    payload = SubplanPayload(**inputs)
except Exception as exc:  # catches everything
    return {"success": False, "error": str(exc)}

This catches ValidationError (intended) but also TypeError, AttributeError, KeyError, and any other exception that might indicate a bug in the validation code itself. These bugs would silently become user-facing error messages instead of surfacing as real failures.

Fix: Catch pydantic.ValidationError and ValueError explicitly.

## Code Review — Bugs & Security Findings Reviewed commits `b888afab` (feat) and `131be21e` (fix) on `feature/m5-subplan-actor` against issue #198 acceptance criteria and `docs/specification.md`. --- ### BUGS #### B1 [High] `parent_decision_id` accepted but never returned to caller **File:** `src/cleveragents/tool/builtins/subplan_tool.py:268-281` The `parent_decision_id` field is accepted in `SubplanPayload` (line 149), stored in the `Decision` object (line 255), but the handler's return dict (lines 268-281) **omits it entirely**. The calling actor has no way to confirm the parent linkage was recorded. The corresponding Behave scenario ("parent_decision_id is included in the decision when provided" at `features/plan_subplan_tool.feature:112-116`) only asserts `success=True` and `decision_type` — it never verifies `parent_decision_id` is present in the result, making the test name misleading and the assertion vacuous. **Fix:** Add `"parent_decision_id": payload.parent_decision_id` to the return dict at line 281, and update the test to assert its value. --- #### B2 [Medium] "Verify all examples are documented" scenario not updated **File:** `features/actor_examples.feature:665-677` The fix commit (`131be21e`) correctly updated the "Verify all example files exist" scenario (line 654) to expect 7 files including `strategy_with_subplan.yaml`. However, the adjacent "Verify all examples are documented" scenario (line 665) was **not** updated to check that `strategy_with_subplan.yaml` is referenced in `docs/reference/actors_examples.md`. Additionally, `docs/reference/actors_examples.md` does not appear to have been updated to reference the new example. --- #### B3 [Medium] Specification naming discrepancy **Spec:** `docs/specification.md` §Child Plan Spawning Mechanism: > *"The `local/create-subplan` tool is independently registered and referenced by name in the actor graph node."* **Implementation:** `builtin/plan-subplan` (`subplan_tool.py:76`) These differ in both namespace (`local/` vs `builtin/`) and name (`create-subplan` vs `plan-subplan`). Either the specification or the implementation should be updated for consistency. --- ### SECURITY #### S1 [High] No ULID format validation on `plan_id`, `dependencies`, `parent_decision_id` **File:** `src/cleveragents/tool/builtins/subplan_tool.py:115,141-143,149-151` `SubplanPayload` accepts arbitrary strings for `plan_id` (only `min_length=1`), `dependencies` (no validation), and `parent_decision_id` (no validation). These fields semantically represent ULIDs. The `Decision` model **in the same codebase** (`domain/models/core/decision.py`) rigorously validates ULID fields with a regex pattern (`^[0-9A-HJKMNP-TV-Z]{26}$`) and dedicated `@field_validator` methods. This inconsistency means: - Malformed `plan_id` values pass the tool handler, then fail inside `Decision(**decision_kwargs)`, producing a confusing error message about the Decision model rather than about the tool input. - Malformed `dependencies` entries flow into the rationale text completely unchecked. **Fix:** Add `@field_validator` methods on `plan_id`, `dependencies`, and `parent_decision_id` in `SubplanPayload` enforcing the same ULID regex pattern used by the `Decision` model. --- #### S2 [Medium] No path validation on `resource_scopes` **File:** `src/cleveragents/tool/builtins/subplan_tool.py:119-121` `resource_scopes` accepts arbitrary strings with no format or sanitization checks. Values like `[""]`, `["../../etc/passwd"]`, or extremely long strings are accepted and flow into the Decision rationale and downstream processing. Early validation (non-empty strings, no path traversal sequences) would prevent garbage from entering the decision tree. --- #### S3 [Low] Overly broad exception handling masks programming errors **File:** `src/cleveragents/tool/builtins/subplan_tool.py:226` ```python try: payload = SubplanPayload(**inputs) except Exception as exc: # catches everything return {"success": False, "error": str(exc)} ``` This catches `ValidationError` (intended) but also `TypeError`, `AttributeError`, `KeyError`, and any other exception that might indicate a bug in the validation code itself. These bugs would silently become user-facing error messages instead of surfacing as real failures. **Fix:** Catch `pydantic.ValidationError` and `ValueError` explicitly.

freemo added the

Points

8

Priority

Medium

State

Verified

labels 2026-03-01 04:36:09 +00:00

aditya force-pushed feature/m5-subplan-actor from cc72ec0936 to eb4e6f59d9 2026-03-02 09:06:37 +00:00 Compare

aditya added 1 commit 2026-03-02 12:37:52 +00:00

Merge branch 'master' into feature/m5-subplan-actor d2f9e4a43b

hamza.khyari commented 2026-03-03 14:01:22 +00:00

Member

Copy Link

Copy Source

P1 — Must Fix

P1-1: Sham rationale test — asserts non-empty, not goal presence

features/steps/plan_subplan_tool_steps.py:302-305

The feature scenario "Rationale is auto-generated for serial subplan" claims to verify the goal is mentioned in the rationale, but the step step_then_rationale_goal only asserts assert rationale (truthy / non-empty). Any non-empty string passes.

@then("the plan_subplan result rationale should mention the goal")
def step_then_rationale_goal(ctx: Context) -> None:
    r = ctx.result
    rationale = r.get("rationale") or ""
    assert rationale, "Expected non-empty rationale"

Fix: Assert the actual goal text appears in the rationale. The When step for this scenario passes goal="Refactor models", so:

def step_then_rationale_goal(ctx: Context) -> None:
    r = ctx.result
    rationale = r.get("rationale") or ""
    goal = r.get("goal") or ""
    assert goal and goal in rationale, (
        f"Expected rationale to contain goal {goal!r}, got: {rationale!r}"
    )

P1-2: source="builtin" silently discarded — ToolSpec has no source field

src/cleveragents/tool/builtins/subplan_tool.py:427

return ToolSpec(
    ...
    source="builtin",  # <-- silently ignored
)

ToolSpec (defined in tool/runtime.py:39) has no source field. Pydantic v2 defaults to extra="ignore", so this kwarg is silently dropped. Any downstream code that tries to inspect spec.source will get AttributeError. If the intent is to tag built-in tools by origin, source must be added to ToolSpec, or this parameter should be removed.

Fix: Either add source: str = Field(default="custom") to ToolSpec, or remove the source="builtin" kwarg.

---

P2 — Should Fix

P2-1: side_effects value diverges from spec

src/cleveragents/tool/builtins/subplan_tool.py:424

The spec (specification.md:18101) defines side_effects: [spawn_subplan] for the create-subplan tool. The implementation uses ["emit_decision"]. While the field is free-form, using a consistent vocabulary with the spec aids auditability and tool-capability queries.

side_effects=["emit_decision"],  # spec says: [spawn_subplan]

Fix: Use ["spawn_subplan"] or ["spawn_subplan", "emit_decision"] to align with the spec vocabulary.

P2-2: Dead TYPE_CHECKING import guard

src/cleveragents/tool/builtins/subplan_tool.py:81-82

if TYPE_CHECKING:
    pass

The block was set up but contains only pass. Either add the intended type-only imports or remove the dead guard.

P2-3: Non-deterministic context_view enum order in JSON schema

src/cleveragents/tool/builtins/subplan_tool.py:92-94, 396

_VALID_CONTEXT_VIEWS is a frozenset, whose iteration order is non-deterministic. The JSON schema enum at line 396 unpacks it via [*list(_VALID_CONTEXT_VIEWS), None], producing a different enum order on each Python process restart. This can cause spurious schema-diff noise in CI.

_VALID_CONTEXT_VIEWS: frozenset[str] = frozenset(
    {"strategist", "executor", "reviewer", "full"}
)

Fix: Use a tuple or sorted() to produce a deterministic order:

_VALID_CONTEXT_VIEWS: tuple[str, ...] = (
    "executor", "full", "reviewer", "strategist"
)

---

P3 — Nit

P3-1: Weak assertion on parallel rationale scenario

features/plan_subplan_tool.feature:98-101

The scenario "Rationale for parallel subplan mentions parallel execution" only asserts that the rationale contains "parallel". It doesn't verify success=True or decision_type="subplan_parallel_spawn". Other parallel scenarios do check these, so this isn't a gap in coverage, but the scenario itself is under-asserted for its title.

P3-2: Subplan payload fields not structurally persisted on Decision

src/cleveragents/tool/builtins/subplan_tool.py:298-309

The Decision object does not carry resource_scopes, merge_strategy, max_parallel, dependencies, or context_view in structured fields — they only appear in the free-text rationale. This means the plan explain tree can display them (via rationale text) but they're not queryable for programmatic use. Worth a follow-up ticket if structured querying is needed.

---

Summary

Severity	Count
P1 (Must Fix)	2
P2 (Should Fix)	3
P3 (Nit)	2

### P1 — Must Fix **P1-1: Sham rationale test — asserts non-empty, not goal presence** `features/steps/plan_subplan_tool_steps.py:302-305` The feature scenario "Rationale is auto-generated for serial subplan" claims to verify the goal is mentioned in the rationale, but the step `step_then_rationale_goal` only asserts `assert rationale` (truthy / non-empty). Any non-empty string passes. ```python @then("the plan_subplan result rationale should mention the goal") def step_then_rationale_goal(ctx: Context) -> None: r = ctx.result rationale = r.get("rationale") or "" assert rationale, "Expected non-empty rationale" ``` **Fix:** Assert the actual goal text appears in the rationale. The `When` step for this scenario passes `goal="Refactor models"`, so: ```python def step_then_rationale_goal(ctx: Context) -> None: r = ctx.result rationale = r.get("rationale") or "" goal = r.get("goal") or "" assert goal and goal in rationale, ( f"Expected rationale to contain goal {goal!r}, got: {rationale!r}" ) ``` **P1-2: `source="builtin"` silently discarded — `ToolSpec` has no `source` field** `src/cleveragents/tool/builtins/subplan_tool.py:427` ```python return ToolSpec( ... source="builtin", # <-- silently ignored ) ``` `ToolSpec` (defined in `tool/runtime.py:39`) has no `source` field. Pydantic v2 defaults to `extra="ignore"`, so this kwarg is silently dropped. Any downstream code that tries to inspect `spec.source` will get `AttributeError`. If the intent is to tag built-in tools by origin, `source` must be added to `ToolSpec`, or this parameter should be removed. **Fix:** Either add `source: str = Field(default="custom")` to `ToolSpec`, or remove the `source="builtin"` kwarg. --- ### P2 — Should Fix **P2-1: `side_effects` value diverges from spec** `src/cleveragents/tool/builtins/subplan_tool.py:424` The spec (`specification.md:18101`) defines `side_effects: [spawn_subplan]` for the create-subplan tool. The implementation uses `["emit_decision"]`. While the field is free-form, using a consistent vocabulary with the spec aids auditability and tool-capability queries. ```python side_effects=["emit_decision"], # spec says: [spawn_subplan] ``` **Fix:** Use `["spawn_subplan"]` or `["spawn_subplan", "emit_decision"]` to align with the spec vocabulary. **P2-2: Dead `TYPE_CHECKING` import guard** `src/cleveragents/tool/builtins/subplan_tool.py:81-82` ```python if TYPE_CHECKING: pass ``` The block was set up but contains only `pass`. Either add the intended type-only imports or remove the dead guard. **P2-3: Non-deterministic `context_view` enum order in JSON schema** `src/cleveragents/tool/builtins/subplan_tool.py:92-94, 396` `_VALID_CONTEXT_VIEWS` is a `frozenset`, whose iteration order is non-deterministic. The JSON schema `enum` at line 396 unpacks it via `[*list(_VALID_CONTEXT_VIEWS), None]`, producing a different enum order on each Python process restart. This can cause spurious schema-diff noise in CI. ```python _VALID_CONTEXT_VIEWS: frozenset[str] = frozenset( {"strategist", "executor", "reviewer", "full"} ) ``` **Fix:** Use a `tuple` or `sorted()` to produce a deterministic order: ```python _VALID_CONTEXT_VIEWS: tuple[str, ...] = ( "executor", "full", "reviewer", "strategist" ) ``` --- ### P3 — Nit **P3-1: Weak assertion on parallel rationale scenario** `features/plan_subplan_tool.feature:98-101` The scenario "Rationale for parallel subplan mentions parallel execution" only asserts that the rationale contains `"parallel"`. It doesn't verify `success=True` or `decision_type="subplan_parallel_spawn"`. Other parallel scenarios do check these, so this isn't a gap in coverage, but the scenario itself is under-asserted for its title. **P3-2: Subplan payload fields not structurally persisted on Decision** `src/cleveragents/tool/builtins/subplan_tool.py:298-309` The `Decision` object does not carry `resource_scopes`, `merge_strategy`, `max_parallel`, `dependencies`, or `context_view` in structured fields — they only appear in the free-text `rationale`. This means the `plan explain` tree can display them (via rationale text) but they're not queryable for programmatic use. Worth a follow-up ticket if structured querying is needed. --- ### Summary | Severity | Count | |---|---| | P1 (Must Fix) | 2 | | P2 (Should Fix) | 3 | | P3 (Nit) | 2 |

aditya added 1 commit 2026-03-03 14:43:52 +00:00

fix(actor): address second round code review on plan_subplan tool ... c4cda5d6d8

- Fix rationale test to assert goal text appears in rationale rather than
  only checking non-empty (P1-1)
- Align side_effects value to specification: use ["spawn_subplan"] instead
  of ["emit_decision"] per docs/specification.md §18259 (P2-1)
- Remove dead TYPE_CHECKING import guard that contained only pass (P2-2)
- Replace frozenset with tuple for _VALID_CONTEXT_VIEWS to ensure
  deterministic JSON schema enum ordering and eliminate CI noise (P2-3)
- Simplify validation error message and schema unpacking now that
  context_view values are in a pre-sorted tuple

aditya added 1 commit 2026-03-03 14:57:58 +00:00

Merge branch 'master' into feature/m5-subplan-actor ... 5f0bc97d46

# Conflicts:
#	CHANGELOG.md

hamza.khyari approved these changes 2026-03-03 17:35:49 +00:00

Dismissed

freemo commented 2026-03-04 01:04:35 +00:00

Owner

Copy Link

Copy Source

Closing as duplicate of #198 (same title, same scope). #198 is assigned to @aditya with active branch feature/m5-subplan-actor.

Closing as duplicate of #198 (same title, same scope). #198 is assigned to @aditya with active branch `feature/m5-subplan-actor`.

freemo closed this pull request 2026-03-04 01:04:35 +00:00

aditya reopened this pull request 2026-03-05 06:29:04 +00:00

aditya added 1 commit 2026-03-05 06:37:23 +00:00

Merge branch 'master' into feature/m5-subplan-actor 7ab5d8661a

aditya dismissed hamza.khyari's review 2026-03-05 06:37:23 +00:00

Reason:

New commits pushed, approval review dismissed automatically according to repository settings

aditya commented 2026-03-05 07:09:17 +00:00

Author

Member

Copy Link

Copy Source

Not duplicate. This is THE #198 PR. I'm assigned @aditya, on feature/m5-subplan-actor
Already approved.

Not duplicate. This is THE #198 PR. I'm assigned @aditya, on `feature/m5-subplan-actor` Already approved.

aditya merged commit fff6bb3833 into master 2026-03-05 07:12:22 +00:00

aditya deleted branch feature/m5-subplan-actor 2026-03-05 07:12:22 +00:00

aditya referenced this issue from a commit 2026-03-05 07:12:23 +00:00

Merge pull request 'feat(actor): add plan_subplan tool and decision emission' (#463) from feature/m5-subplan-actor into master

Sign in to join this conversation.

Reviewers

No Reviewers

CoreRasurae

Labels

Clear labels

auto/needs-reevaluation

Controller deferred this PR; awaiting Phase 6+ scope-evaluator or operator re-enablement.

controller-managed

Auto-agents controller manages this PR/issue (see tools/controller/deploy/RUNBOOK.md). Remove this label to abandon controller management.

overdue

auto/blocked-by-deps

PR blocked by an open issue dependency. Operator must close the dep (or remove the dependency link) before the merge driver can act. Auto-cleared by merge_drive when no open deps remain.

auto/ci-timeout

Most recent merge cycle hit CI timeout. Driver excludes this PR while last merge_cycle row is < 30 min old; label persists thereafter as visible history.

auto/claimed-implementer

Currently being processed by an implementer worker.

auto/claimed-merge

Currently being processed by the merge driver.

auto/claimed-reviewer

Currently being processed by a reviewer worker.

auto/driver-down

Merge driver heartbeat stale; pipeline halted. Closed automatically on next clean tick.

auto/invariant-violation

Detected master commit violating the strict merge invariant. Tracked as an issue (not a PR label); kept here for label completeness.

auto/last-attempt-tier-0

In-cycle escalation: most recent attempt ran at the Tier 0 slot (`tier-0`). Slot's model defined in .opencode/models/tiers.yaml.

auto/last-attempt-tier-1

In-cycle escalation: most recent attempt ran at the Tier 1 slot (`tier-1`). Slot's model defined in .opencode/models/tiers.yaml.

auto/last-attempt-tier-2

In-cycle escalation: most recent attempt ran at the Tier 2 slot (`tier-2`). Slot's model defined in .opencode/models/tiers.yaml. Gated behind IMPLEMENTER_ESCALATION_TIER2_ENABLED.

auto/last-attempt-tier-min

In-cycle escalation: most recent attempt ran at the Tier -1 slot (`tier-min`). Slot's model defined in .opencode/models/tiers.yaml. Suffix is ``-min`` (not ``--1``) so the Forgejo UI reads naturally.

Automation Tracking

Tracking issues used by the AI Automation system for agents to communicate and report.

auto/needs-conflict-resolution

Rebase conflict needs LLM conflict-resolver.

auto/needs-implementer

Failing CI needs implementer attention.

auto/postmortem

Documenting a driver incident or rollback.

auto/ready-to-merge

Reviewer has APPROVED this PR and no later REQUEST_CHANGES is outstanding. The merge driver requires this label to even consider a PR for merging. Set by the reviewer worker on APPROVE; cleared on REQUEST_CHANGES.

auto/restart-throttled

Train repeatedly lost master-tempo races. Driver excludes via merge_cycle until cooldown elapses; label persists as visible history.

auto/revert

Revert PR backing out an invariant violation. Fast-tracked through the merge driver.

auto/sentinel

Sentinel PR duplicated from upstream into a personal fork by tools/duplicate_prs_to_fork.py for pipeline testing. Lives only in the fork; the canonical pipeline never sees it.

auto/stale-inactivity

No implementer activity for N days. Flagged for human review. Auto-cleared on next push to head branch.

auto/unstable

Repeatedly fails on current master (>= 3 ci-fail-on-rebased-sha releases in 12 h). Excluded from driver until human triage.

Blocked

A ticket in a blocked state and unable to complete until some other task is completed first.

Bounty

$100

A bounty of $100 for any open-source contributor who provides a MR that solves this issue

Bounty

$1000

A bounty of $1000 for any open-source contributor who provides a MR that solves this issue

Bounty

$10000

A bounty of $10000 for any open-source contributor who provides a MR that solves this issue

Bounty

$20

A bounty of $20 for any open-source contributor who provides a MR that solves this issue

Bounty

$2000

A bounty of $2000 for any open-source contributor who provides a MR that solves this issue

Bounty

$250

A bounty of $250 for any open-source contributor who provides a MR that solves this issue

Bounty

$50

A bounty of $50 for any open-source contributor who provides a MR that solves this issue

Bounty

$500

A bounty of $500 for any open-source contributor who provides a MR that solves this issue

Bounty

$5000

A bounty of $5000 for any open-source contributor who provides a MR that solves this issue

Bounty

$750

A bounty of $750 for any open-source contributor who provides a MR that solves this issue

MoSCoW

Could have

Could have feature in order to satisfy the epic/legendary.

MoSCoW

Must have

Must have feature in order to satisfy the epic/legendary.

MoSCoW

Should have

Should have feature in order to satisfy the epic/legendary.

Needs Feedback

There are questions in the ticket that can not be completed until the project owner provides clarity.

Points

1

1 man-hours worth of work for an expert with no learning curve.

Points

13

13 man-hours worth of work for an expert with no learning curve.

Points

2

2 man-hours worth of work for an expert with no learning curve.

Points

21

21 man-hours worth of work for an expert with no learning curve.

Points

3

3 man-hours worth of work for an expert with no learning curve.

Points

34

34 man-hours worth of work for an expert with no learning curve.

Points

5

5 man-hours worth of work for an expert with no learning curve.

Points

55

55 man-hours worth of work for an expert with no learning curve.

Points

8

8 man-hours worth of work for an expert with no learning curve.

Points

88

88 man-hours worth of work for an expert with no learning curve.

Priority

Backlog

This ticket has backlogged priority and is not to be worked on yet

Priority

CI Blocker

Critical priority issue that blocks CI/CD pipeline and prevents PR merges

Priority

Critical

The priority is critical

Priority

High

The priority is high

Priority

Low

The priority is low

Priority

Medium

The priority is medium

Signed-off: Owner

When an epic or legendary is in review it must be signed off by owner, tech lead, and scrum master before being marked as completed.

Signed-off: Scrum Master

When an epic or legendary is in review it must be signed off by owner, tech lead, and scrum master before being marked as completed.

Signed-off: Tech Lead

When an epic or legendary is in review it must be signed off by owner, tech lead, and scrum master before being marked as completed.

Spike

A ticket for learning a tool or technology that is needed to be able to do future planning and design.

State

Completed

The ticket has been fully implemented, completed, and merged with the source code. This label should only be applied once a ticket is closed.

State

Duplicate

A ticket that represents the same content as an existing ticket.

State

In Progress

A ticket that is actively being developed.

State

In Review

A ticket that has had some code completed to implement but is waiting to pass peer review and is not yet merged in.

State

Paused

This ticket's work started but wasn't finished. It's on hold (likely in a feature branch) and will be resumed later, either due to a blocker or a delay.

State

Unverified

All new tickets start in this state. A developer may set it to show the ticket is unverified. This means we haven't agreed to work on it. It will either move to a verified state or be closed as wontdo.

State

Verified

The issue has been verified by a developer as legitimate. It will be worked on and verified tickets are now considered part of the backlog.

State

Wont Do

This ticket has been decided it wont be done. This may mean the bug has been determined to not be real (cant verify) or the feature is one we have decided we dont want to adopt.

Type

Automation

Any edits or discussion about the AI automated coding system.

Type

Bug

Something that doesnt work as intended.

Type

Discussion

Anytime a ticket represents a discussion about a subject and doesnt fall into one of the other categories.

Type

Documentation

An error or improvement needed in the documentation.

Type

Epic

Any first tier epic. That is, an epic which contains only issues as children and will not have sub-epics.

Type

Feature

Some new functionality not present.

Type

Legendary

A type of Epic which will contain other Epics.

Type

Refactor

A code change that restructures existing code without changing its external behavior.

Type

Support

Someone needs help using the project.

Type

Task

A generic task that doesnt fit into the other type categories.

Type

Testing

Work exclusively focusing on fixing or expanding testing.

No Label

Points

8

Priority

Medium

State

Verified

Type

Feature

Milestone

No items

No Milestone v3.4.0

Projects

Clear projects

No project

Assignees

Clear assignees

aditya (Aditya Chhabra) aleenaumair (Aleena Umair) brent.edwards (Brent Edwards) CoreRasurae (Luis Mendes) drew (Drew Morris) eugen.thaci (Eugen Thaci) freemo (Jeffrey Phillips Freeman) HAL9000 (HAL 9000) HAL9001 (HAL9001) hamza.khyari (Hamza Khyari) hurui200320 (Rui Hu) justin.morris khird (Kyle Hird) org.cleveragents

No Assignees

4 Participants

Notifications

Subscribe

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: cleveragents/cleveragents-core#463