subagent-output-templating

Sub-Agent Output Templating

Overview

This skill provides standardized templates for sub-agent OUTPUT formatting. It complements subagent-prompting (P0.1) which defines INPUT structure:

Skill	Purpose
subagent-prompting	How to prompt sub-agents (GOAL/CONSTRAINTS/CONTEXT/OUTPUT)
subagent-output-templating	How sub-agents report results (logs, summaries, diagnostics)

Use this skill when:

• Defining output requirements for sub-agent invocations
• Parsing sub-agent results in pipeline stages
• Ensuring consistent log formats across all agents

---

Log File Format

File Location

logs/{agent-name}-{YYYYMMDD-HHMMSS}.yaml

Example: logs/bulwark-code-auditor-20260111-143022.yaml

YAML Schema

# Required: Metadata block
metadata:
  agent: {agent-name}           # e.g., bulwark-code-auditor
  timestamp: {ISO-8601}         # e.g., 2026-01-11T14:30:22Z
  model: {model-used}           # sonnet, haiku, or opus
  task_id: {unique-identifier}  # For tracking across pipeline stages
  duration_ms: {execution-time} # Execution duration in milliseconds

# Required: Goal from the prompt (for traceability)
goal: "{GOAL from 4-part prompt}"

# Required: Completion report
completion:
  why:
    problem: "{What was broken/missing}"
    root_cause: "{Why it happened}"
    solution: "{What was implemented}"

  what:
    - file: {path}
      lines: "{range}"
      change: "{description}"

  trade_offs:
    gained:
      - "{benefit 1}"
    cost:
      - "{drawback 1}"

  risks:
    - risk: "{description}"
      mitigation: "{how addressed}"
      severity: {low|medium|high|critical}

  next_steps:
    - "{action item 1}"

# Required for code-writing agents (omit for read-only agents):
# Pipeline suggestions from implementer-quality.sh output
pipeline_suggestions:
  - pipeline: "{recommended pipeline name}"
    target_files:
      - "{file path}"
    reason: "{why this pipeline is recommended}"

# Required: Summary for main thread (100-300 tokens)
summary: |
  {Concise summary for main thread consumption}

# Required: Diagnostic output
diagnostics:
  model_requested: {model}
  model_actual: {model}
  context_type: {main|forked}
  parent_vars_accessible: {true|false}
  hooks_fired:
    - {hook-name}
  execution_time_ms: {duration}
  completion_status: {success|error|timeout}