AgentSkillsCN

report-template

将Scout阶段的各项分析成果(构建检查器、运行时检查器、Git取证、概念建模)综合汇总,形成一份可供决策参考的系统风险报告。

SKILL.md
--- frontmatter
name: report-template
description: Synthesize all Scout-phase analyses (build-inspector, runtime-inspector, git-forensics, concept-modeler) into a decision-ready system risk report.

The Synthesizer’s Manual

“Data is not information. Information is not knowledge. Knowledge is not wisdom.” — T.S. Eliot

Your goal is to transform raw analytical data into architectural wisdom that can drive decisions.


⚠️ Mandatory Deep Thinking

[!IMPORTANT] Before generating the report, you must invoke the mcp_sequential-thinking_sequentialthinking tool and reason for 5–10 steps, as appropriate.

Example thinking prompts:

  1. “Do the build boundaries discovered by build-inspector align with the IPC boundaries found by runtime-inspector?”
  2. “Do the high-coupling file pairs identified by git-forensics cross build boundaries?”
  3. “Are the missing components identified by concept-modeler correlated with observed risks?”
  4. “Which findings are 🔴 blocking risks versus 🟡 risks to monitor?”

⚡ Quick Start

  1. Load the template (MANDATORY) Use view_file references/REPORT_TEMPLATE.md. Your report must strictly follow this structure.

  2. Aggregate all findings from:

    • build-inspector → Build Roots, Topology
    • runtime-inspector → IPC Surfaces, Contract Status
    • git-forensics → Coupling Pairs, Hotspots
    • concept-modeler → Entities, Missing Components
  3. Draft the report Use sequentialthinking to organize logical connections and conclusions.

  4. Publish (CRITICAL) You must use write_to_file to save the report to:

    code
    scout/SCOUT_RISK_REPORT.md
    

    Do not output the report only in chat.


✅ Completion Checklist

Before proceeding to the next phase, verify:

  • Output file created: scout/SCOUT_RISK_REPORT.md
  • Includes: System Fingerprint, Component Map, Risk Matrix, Feature Landing Guide
  • User has confirmed the findings

🛠️ Synthesis Ritual

1. Executive Summary

  • Elevator pitch: Describe system health in 30 seconds.
  • Focus areas: Technical debt, critical risks, reliability.

2. Dark Matter Detection

  • Do not only list what exists. Explicitly list what is missing.
  • Checklist: Logging? Error handling? CI/CD? Secret management? Version handshakes?

3. Cross-Verification

  • Does build-inspector claim “unified workspace management”?
  • Does git-forensics reveal “high coupling across build roots”?
  • Conclusion: Hidden logical coupling detected → refactoring candidate.

4. Human Checkpoint

  • Force user confirmation: “Is this report complete?”
  • No Blueprint phase allowed until this report is signed off.

🛡️ Master Rules

  1. No hallucinations: Every claim must be traceable to a source file.
  2. Brutal honesty: Be direct. If it’s a mess, say it’s a mess.
  3. Action-oriented: Every issue listed must imply a path forward (refactor / rewrite / accept).

🧰 Toolbox

  • references/REPORT_TEMPLATE.md — primary report template

Consumers

The direct consumers of this report in the /blueprint phase are:

  • System Architect — relies on your risk list to design mitigation strategies
  • Complexity Guard — uses your findings to audit RFC and design complexity

The quality of your analysis directly determines the quality of the next design phase.