long-task-quality

name	long-task-quality
description	Use after TDD cycle in a long-task project - enforces coverage gate, mutation gate, and fresh verification evidence before marking features as passing

Quality Gates — SubAgent Dispatch

Delegate quality gate execution to a SubAgent with fresh context. The main Agent only dispatches and parses the structured result — it never reads coverage reports, mutation output, or test runner output directly.

Announce at start: "I'm using the long-task-quality skill to run quality gates via SubAgent."

Step 1: Construct SubAgent Prompt

Build the prompt from current session state. Do NOT read any source code, test output, or coverage reports yourself.

You are a Quality Gates execution SubAgent.

## Your Task
1. Read the execution rules: Read {skills_root}/long-task-quality/references/quality-execution.md
2. Read long-task-guide.md in the project root for test/coverage/mutation commands and environment activation
3. Execute all 4 gates in order (Gate 0 → 1 → 2 → 3)
   - **Note**: Static analysis tools (Design §13.4) are enforced during TDD Refactor, not here. If Design doc §13.7 documents code generation directories, exclude them from coverage measurement in Gate 1.
4. If a gate fails, fix and retry per the rules (max 3 attempts per gate)
5. Return your result using the Structured Return Contract at the end of the execution rules

## Input Parameters
- Feature ID: {feature_id}
- Feature: {feature_json}
- quality_gates thresholds: {quality_gates_json}
- tech_stack: {tech_stack_json}
- Working directory: {working_dir}
- Feature test files: {feature_test_files}  (test files written/modified during TDD for this feature — used for mutation_feature scoping)
- Active feature count: {active_feature_count}  (total non-deprecated features — compared against mutation_full_threshold to decide mutation scope)

## Key Constraint
- Do NOT mark the feature as "passing" in feature-list.json — only report results
- If a tool/environment error cannot be resolved after 1 retry, set Verdict to BLOCKED

Replace {skills_root} with the path to the skills directory (e.g., skills in the project or the installed plugin path).

Step 2: Dispatch SubAgent

Claude Code: Use the Agent tool:

Agent(
  description = "Quality Gates for feature #{feature_id}",
  prompt = [the constructed prompt above]
)

OpenCode: Use @mention syntax or the platform's native subagent mechanism with the same prompt content.

Step 3: Parse Result

Read the SubAgent's returned text and locate the ### Verdict: line:

• ### Verdict: PASS

  1. Extract Metrics table (coverage %, mutation score)
  2. Extract Next Step Inputs (coverage_line, coverage_branch, mutation_score)
  3. Record in task-progress.md: "Quality Gates: PASS (line {X}%, branch {Y}%, mutation {Z}%)"
  4. Proceed to next step (Feature-ST)

• ### Verdict: FAIL

  1. Read the Issues table — identify which gate failed and why
  2. If the SubAgent already attempted fixes (per the 3-retry rule), escalate to user via AskUserQuestion with the failure details
  3. If fixable by re-dispatching (e.g., environment issue resolved), construct a new prompt and dispatch again (max 3 total dispatches)

• ### Verdict: BLOCKED

  1. Read the Issues table — identify the blocker (tool not installed, environment error, etc.)
  2. Escalate to user via AskUserQuestion with the blocker details and what was attempted

Integration

Called by: long-task-work (Step 8) Requires: TDD cycle completed (long-task-tdd passed — tests exist and pass) Produces: Structured summary (coverage %, mutation %, per-gate pass/fail) Chains to: long-task-feature-st (via Work Step 9)