ssaattww/backend-task-quality-loop

Backend Task Quality Loop

Purpose

• Execute /task, /build, and /review equivalent behavior directly.
• Maintain deterministic task-by-task quality execution.

Execution Contract

This skill follows the non-entry execution contract standard. Required contract_extensions: execution_mode. See codex-execution-contract.md and non-entry-execution-contract-template.md for full rules.

input:
  objective: "Run autonomous task-quality cycle"
  contract_extensions: { execution_mode: "autonomous-loop" }
output:
  status: "completed"
  quality_gate:
    gate_id: "backend-task-quality-check"
    gate_type: "implementation"
    trigger: "post-cycle validation"
    criteria:
      - "Task-quality checks are complete"
      - "Execution mode remains aligned with the contract"
    result: "pass"
    evidence:
      - "Task quality gate passed"
    blockers: []
    branching:
      on_pass: "handoff"
      on_fail: "revise"
      max_cycles: 2
  contract_extensions: { execution_mode: "autonomous-loop" }

Contract Compliance

• Emit structured output compliant with codex-execution-contract.md.
• Always include baseline output fields: status, summary, changed_files, tests, quality_gate, blockers, next_actions.
• Validate required input fields from ../workflow-entry/references/non-entry-execution-contract-template.md (objective, scope, constraints, acceptance_criteria, allowed_commands, sandbox_mode) before proceeding.
• Echo required skill extensions in contract_extensions: execution_mode.
• Treat missing required fields as contract violations and regenerate output before handoff.
• On contract violation (missing/invalid field, invalid status value, or missing extension keys): do not proceed; emit status: blocked with violation description in blockers.
• Reference: non-entry-execution-contract-template.md.

Execution Modes

• single-task (/task equivalent)
• autonomous-loop (/build equivalent)
• compliance-review (/review equivalent)

Mandatory Per-Task Cycle

1. Define a single atomic task unit.
2. Implement with coding-principles.
3. Run escalation check (scope/risk/blockers).
4. Run quality gate with ai-development-guide.
5. Validate tests (testing-principles).
6. Mark ready for commit/report.

Quality Gate Minimum

• Format/style checks
• Lint/static checks
• Build/compile
• Unit tests
• Integration tests when impacted

Autonomous Loop Rules

• Execute one task unit at a time.
• If task files are missing but a plan exists, generate task units first.
• On requirement changes, emit [Stop: requirement-change-detected] + [Approve: route-selection].
• Never defer quality checks to the end of the loop.

Compliance Review Mode

1. Compare implementation against Design Doc acceptance criteria.
2. Score coverage and list gaps.
3. Before applying fixes, emit [Stop: pre-implementation-approval] + [Approve: implementation-start] unless already approved for current scope.
4. Re-run full quality gate.
5. Re-score and report remaining non-fixable issues.

Hard Rules

• No multi-task batching in one quality cycle.
• No silent fallback to hide errors.
• No completion claim without test/build evidence.

Quality Gate Evidence

• This executor owns quality_gate emission and branching using quality-gate-evidence-template.md.
• Emit canonical fields: gate_id, gate_type, trigger, criteria, result, evidence, blockers, branching.
• Use gate_type: implementation for backend task/build/review execution loops.
• Normalize local statuses into result: pass|fail|blocked before handoff.
• If result: blocked, emit [Stop: quality-gate-failed] and pause for escalation handling.

Stop/Approval Protocol

Use canonical markers: [Stop: <Gate Name>]. Classify every stop as approval_gate or escalation_gate. At each stop, emit a full gate record: gate_name, gate_type, trigger, ask_method, required_user_action, resume_if, fallback_if_rejected. Default ask_method is AskUserQuestion. Resume an approval_gate only with explicit user approved: true; resume an escalation_gate only after user direction or reroute. Respect batch boundary: no autonomous implementation starts before [Stop: pre-implementation-approval] is approved. Enforce max_revision_cycles: 2; overflow requires human intervention. Agent-local review scores never replace user approvals.

Stop points for this skill:

• [Stop: pre-implementation-approval] (approval_gate)
• [Stop: high-risk-change] (approval_gate)
• [Stop: requirement-change-detected] (escalation_gate)
• [Stop: quality-gate-failed] (escalation_gate)
• [Stop: revision-limit-reached] (escalation_gate)

Full protocol and payload schema: ../workflow-entry/references/stop-approval-section-template.md.