Run Parity Test B

run-parity-test-b — Entry Skill

> Entry point for the parity-test-b pipeline. Orchestrates the sequential Task()-based dispatch of the analyzer, reviewer, and reporter agents on Tier 1 (Claude Code). The reviewer runs with worktree isolation + permissionMode: plan + disallowedTools: Write, Edit, Bash — structural write/review isolation.

This skill initializes the pipeline run, resolves the scope root and run ID, dispatches the analyzer, reviewer (structural isolation), and reporter agents in sequence via the native_task mechanism, and applies the cleanup contract (C20) on completion. The reviewer cannot write files — its verdict is read from terminal output text.

Platform Context

• Tier: tier_1 (Claude Code)
• Dispatch mechanism: native_task — DISPATCH(mode="task", agent="{name}", isolation="worktree", context={...}).
• modelfieldformat: shorthand — model_tier in each agent's YAML frontmatter; resolver maps to concrete model.
• Reviewer isolation: structural — permissionMode: plan + disallowedTools: Write, Edit, Bash + separate worktree.
• Degradation warnings: None for Tier 1.

Workflow

PHASE 0: PREFLIGHT

1. Resolve {ROOT} via sk-pipeline-paths.
2. Resolve {runId} = ISO-8601 compact timestamp.
3. Create temp directory: {ROOT}/superpipelines/temp/parity-test-b/{runId}/.
4. Create output/ directory if absent: {ROOT}/output/.
5. If the user has not supplied the path to the input JSON file, ask for it now. Record as {INPUT_PATH}.
6. Initialize pipeline-state.json at {ROOT}/superpipelines/temp/parity-test-b/{runId}/pipeline-state.json:

{
  "pipeline_id": "parity-test-b",
  "run_id": "{runId}",
  "started_at": "{iso8601}",
  "plugin_version": "2.0.0",
  "pattern": "1",
  "status": "running",
  "current_phase": 0,
  "metadata": {
    "source_tier": "tier_1",
    "runtime_tier": "tier_1",
    "model_field_format": "shorthand",
    "resolved_models": {
      "analyzer": "claude-haiku-4-5-20251001",
      "reviewer": "claude-sonnet-4-6",
      "reporter": "claude-haiku-4-5-20251001"
    }
  },
  "phases": [
    {
      "index": 0,
      "step_id": "analyzer",
      "name": "analyze",
      "status": "pending",
      "agent": "agents/superpipelines/parity-test-b/analyzer.md",
      "outputs": [],
      "error": null
    },
    {
      "index": 1,
      "step_id": "reviewer",
      "name": "review",
      "status": "pending",
      "agent": "agents/superpipelines/parity-test-b/reviewer.md",
      "reviewer_isolation": "structural",
      "outputs": [],
      "error": null
    },
    {
      "index": 2,
      "step_id": "reporter",
      "name": "report",
      "status": "pending",
      "agent": "agents/superpipelines/parity-test-b/reporter.md",
      "outputs": [],
      "error": null
    }
  ]
}

PHASE 1: DISPATCH ANALYZER (native_task)

DISPATCH(
  mode="task",
  agent="analyzer",
  isolation="worktree",
  context={
    "input_path": "{INPUT_PATH}",
    "findings_output_path": "{ROOT}/superpipelines/temp/parity-test-b/{runId}/findings.json",
    "state_path": "{ROOT}/superpipelines/temp/parity-test-b/{runId}/pipeline-state.json",
    "run_id": "{runId}",
    "root": "{ROOT}"
  }
)

Wait for terminal status:

• DONE → update phases[0].status = "completed"; advance to Phase 2.
• DONE_WITH_CONCERNS → update phases[0].status = "completedwithconcerns"; surface concerns; advance to Phase 2.
• NEEDS_CONTEXT → update phases[0].status = "blocked"; top-level status = "blocked"; GO TO CLEANUP (preserve).
• BLOCKED → update phases[0].status = "blocked"; top-level status = "blocked"; GO TO CLEANUP (preserve).

PHASE 2: DISPATCH REVIEWER (native_task — structural isolation)

The reviewer agent has permissionMode: plan and disallowedTools: Write, Edit, Bash. It cannot write files. Capture the REVIEWER VERDICT block from its terminal output text.

DISPATCH(
  mode="task",
  agent="reviewer",
  isolation="worktree",
  context={
    "findings_path": "{ROOT}/superpipelines/temp/parity-test-b/{runId}/findings.json",
    "state_path": "{ROOT}/superpipelines/temp/parity-test-b/{runId}/pipeline-state.json",
    "run_id": "{runId}",
    "root": "{ROOT}"
  }
)

After the reviewer completes:

1. Extract the REVIEWER VERDICT block from terminal output text. Record as {REVIEWER_OUTPUT}.
2. Parse verdict from the block: approved, approved_with_concerns, or rejected.
3. Update pipeline-state.json:

• phases[1].outputs → [{ "verdict": "{verdict}", "reviewer_notes": "{REVIEWER_OUTPUT}" }]

Wait for terminal status:

• DONE (verdict: approved) → update phases[1].status = "completed"; advance to Phase 3.
• DONE_WITH_CONCERNS (verdict: approvedwithconcerns) → update phases[1].status = "completedwithconcerns"; surface concerns; advance to Phase 3.
• BLOCKED (verdict: rejected) → update phases[1].status = "blocked"; top-level status = "blocked"; surface rejection; GO TO CLEANUP (preserve). DO NOT dispatch reporter.
• NEEDS_CONTEXT → update phases[1].status = "blocked"; GO TO CLEANUP (preserve).

PHASE 3: DISPATCH REPORTER (native_task)

Only dispatched if reviewer verdict is approved or approved_with_concerns.

DISPATCH(
  mode="task",
  agent="reporter",
  isolation="worktree",
  context={
    "findings_path": "{ROOT}/superpipelines/temp/parity-test-b/{runId}/findings.json",
    "reviewer_verdict": "{verdict}",
    "reviewer_notes": "{REVIEWER_OUTPUT}",
    "report_output_path": "{ROOT}/output/parity-test-b-report.md",
    "state_path": "{ROOT}/superpipelines/temp/parity-test-b/{runId}/pipeline-state.json",
    "run_id": "{runId}",
    "root": "{ROOT}"
  }
)

Wait for terminal status:

• DONE → update phases[2].status = "completed"; advance to Phase 4.
• DONE_WITH_CONCERNS → update phases[2].status = "completedwithconcerns"; surface concerns; advance to Phase 4.
• NEEDS_CONTEXT → update phases[2].status = "blocked"; GO TO CLEANUP (preserve).
• BLOCKED → update phases[2].status = "blocked"; GO TO CLEANUP (preserve).

PHASE 4: FINALIZE

1. Update pipeline-state.json:

• status: "completed" (or "completed_with_concerns")
• completed_at: ISO-8601 timestamp

1. Cleanup contract (C20):

• On DONE or DONE_WITH_CONCERNS: write status: completed to pipeline-state.json, then delete {ROOT}/superpipelines/temp/parity-test-b/{runId}/.
• On BLOCKED or NEEDS_CONTEXT: preserve the temp directory.

1. Confirm to the user:

> Pipeline parity-test-b completed. Data quality report written to: {ROOT}/output/parity-test-b-report.md

1. Emit terminal status: DONE (or DONE_WITH_CONCERNS / BLOCKED / NEEDS_CONTEXT as applicable).

• NEVER call Task(subagent_type=...) directly — use abstract DISPATCH(mode="task", ...) notation.
• NEVER hardcode platform paths — always use {ROOT} resolved via sk-pipeline-paths.
• NEVER pass file contents in dispatch context — pass file paths only.
• NEVER dispatch the reporter if the reviewer verdict is rejected or the reviewer emits BLOCKED.
• ALWAYS capture the REVIEWER VERDICT block from the reviewer's terminal output before advancing.
• ALWAYS update pipeline-state.json after each phase completes.
• NEVER advance past a BLOCKED or NEEDS_CONTEXT status without human input.
• ALWAYS apply the C20 cleanup contract: delete temp dir on DONE/DONEWITHCONCERNS, preserve on BLOCKED/NEEDS_CONTEXT.
• Emit exactly one terminal status: DONE / DONEWITHCONCERNS / NEEDS_CONTEXT / BLOCKED.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: gustavo-meilus
• Source: gustavo-meilus/superpipelines
• License: MIT

Install and usage instructions live in the source repository linked above.