82 lines
5.3 KiB
Plaintext
82 lines
5.3 KiB
Plaintext
version: 1
|
|
goal_summary: "Implement a structured three-persona planning pipeline for new task creation and wire it into quality gates that prevent spaghetti-like outcomes."
|
|
steps[6]:
|
|
- id: "step-1"
|
|
title: "Map current goal-planning execution path"
|
|
purpose: "Identify where new tasks are created and where planner output is generated, stored, and consumed."
|
|
notes: This is required to avoid adding persona logic in the wrong layer and breaking existing task creation contracts.
|
|
inputs[1]: "Current new-task creation flow"
|
|
outputs[1]: Execution map of planner entry/egress points
|
|
dependencies[0]:
|
|
verification[1]:
|
|
- label: Baseline understanding
|
|
commands[3]: Trace task creation path from command/API input to persisted task goal,"Locate planner modules, serializers, and validation hooks",Document current persona/prompt hooks (if any)
|
|
cleanup_requirements[0]:
|
|
status: done
|
|
attempts: 1
|
|
- id: "step-2"
|
|
title: Reset stale active state
|
|
purpose: "Re-run finalization of three-persona planner sequencing before execution resumes"
|
|
notes: "Recovered stale active step state required step-2 to be retried before proceeding."
|
|
inputs[0]:
|
|
outputs[0]:
|
|
dependencies[0]:
|
|
verification[0]:
|
|
cleanup_requirements[0]:
|
|
status: done
|
|
attempts: 1
|
|
- id: "step-3"
|
|
title: Implement sequential persona pipeline in goal planner
|
|
purpose: "Apply PO -> Senior Engineer -> Future Engineer transformations and merge outputs into a single maintainability-first goal artifact."
|
|
notes: Recovery shows planner execution was reset; this step is now the active unblocked execution path. Controller recovered this step from stale active state and returned it to todo.
|
|
inputs[1]: Persona contract and config
|
|
outputs[2]: Updated planner pipeline implementation,Deterministic merged goal artifact
|
|
dependencies[1]: "step-2"
|
|
verification[1]:
|
|
- label: Determinism
|
|
commands[3]: Add unit tests asserting fixed output ordering and stable merge for same input seed/state,Add tests for conflict resolution and fallback behavior,... 1 more omitted
|
|
cleanup_requirements[1]{label,description}:
|
|
Safe integration,Remove inline prompt branching that bypasses the new pipeline.
|
|
status: done
|
|
attempts: 1
|
|
- id: "step-4"
|
|
title: "Reopen step-4 implementation"
|
|
purpose: Complete remaining implementation work before downstream validation/rollout tasks.
|
|
notes: "Recovery flagged step-4 as stale-active; it must be treated as unfinished to keep planner gating chain valid."
|
|
inputs[0]:
|
|
outputs[0]:
|
|
dependencies[0]:
|
|
verification[0]:
|
|
cleanup_requirements[0]:
|
|
status: done
|
|
attempts: 1
|
|
- id: "step-5"
|
|
title: Add regression tests and rollout docs
|
|
purpose: "Validate the three-stage PO→Senior Engineer→Architect flow with deterministic templates, explicit constraints, and maintenance-focused rejection/annotation behavior before proceeding."
|
|
notes: "Keep this step as the first blocker: lock in behavioral contracts and anti-bypass coverage, then add concise docs so future work cannot drift into one-off prompts. Controller recovered this step from stale active state and returned it to todo."
|
|
inputs[2]: Implemented pipeline and gates,"Three-stage persona templates and merge rules"
|
|
outputs[3]: Regression tests for staged transformations,"Tests for downgrade/reject/delta-capture logic that flags maintenance-risk solutions",Developer docs for staged goal generation and constraints
|
|
dependencies[1]: "step-4"
|
|
verification[1]:
|
|
- label: Regression safety
|
|
commands[3]: "Run new-task planner unit/integration test set for all 3 passes","Run decision-path tests for reject/downgrade/annotate outcomes","Run existing task-creation and verifier-loop regression tests"
|
|
cleanup_requirements[1]{label,description}:
|
|
Documentation clarity,"Delete stale references to old single-pass planning behavior."
|
|
status: done
|
|
attempts: 1
|
|
- id: "step-6"
|
|
title: "Enable three-stage persona telemetry and rollout guardrails"
|
|
purpose: "Add immutable planner instrumentation and hard enforcement for product-owner, senior-engineer, senior-maintainer stage transitions before execution."
|
|
notes: "Implement stage event emission (`Persona stage`, `stage-1`, `stage-2`, `stage-3`) with deterministic persona-pass metadata; block single-pass/bypass paths so every new-task goal goes through all three stages; keep this guardrail work separate from doc/test-generation output paths."
|
|
inputs[0]:
|
|
outputs[0]:
|
|
dependencies[1]: "step-4"
|
|
verification[2]:
|
|
- label: Stage telemetry contract
|
|
commands[2]: "Emit planner session events for every pass including persona template version, constraints, risks, architectural context, and acceptance criteria","Emit explicit `Persona stage` and `stage-1/2/3` transitions from the same approved pipeline path for all new-task goals"
|
|
- label: Guardrail enforcement
|
|
commands[2]: "Reject execution when single-pass planning or stage skipping is detected","Add rollout counters for rejected/annotated goals by failure type and stage, and alert on threshold breaches"
|
|
cleanup_requirements[1]{label,description}:
|
|
Operational overhead,"Disable debug-level per-task instrumentation in non-debug environments once the new telemetry/guards are stable."
|
|
status: done
|
|
attempts: 1 |