case-review – skill Lập trình cho agent AI

Evidence Graph Review

When to Use

• Before delivering an analysis report, verify traceability and completeness.
• Auditing whether conclusions are backed by recorded evidence.

Use this skill when a reverse engineering, forensics, CTF, or authorized security case needs a defensible handoff. It audits the existing work/<case>/ package without changing the case or touching a target.

Scope

This skill covers:

• Scope metadata and target-activity readiness
• Evidence record structure and reproducibility fields
• References from work items and timeline entries to Evidence
• Structured Findings and Paths in report Markdown
• Optional SHA-256 verification for case-local artifacts
• A Markdown or JSON review result for a report handoff

It MUST NOT perform reconnaissance, exploitation, dynamic instrumentation, or target changes. Those actions belong to the routed analysis skill and require the case scope gate.

Imported scope

This tree bundles SKILL.md only. scripts/review_case.py, skills/scripts/append-evidence.ps1 and routed modules are not included. Obtain a reviewed matching upstream installation before any dependent command; otherwise perform the inline traceability checks manually and report script/hash/strict-mode verification as not run. Script behavior described below is an upstream claim, not a test result for this import.

Tool dependencies

| Tool | Required | Purpose | Auto-bootstrap |

|------|----------|---------|---------------|

| Python 3.9+ | Yes | Runs the read-only case review script | No, use the platform Python installation |

No network access or third-party package is required.

Workflow

Phase 1: Intake

Run the review against the existing case directory:

python3 skills/case-review/scripts/review_case.py work/<case> --format markdown

Confirm that scope.md, timeline.md, workitems.md, and evidence/ are present. A non-strict review reports scope warnings while a strict review treats warnings as handoff blockers.

建议下一步（选一个编号）

1. 修复 scope.md 中的授权、范围或 network_profile 字段

2. 继续检查 Evidence 记录的可复现命令和来源

3. 导出当前 review 结果并附到阶段性报告

4. 换 JSON 输出接入 CI 或其他审查工具

5. 暂停，先确认审查范围

Phase 2: Traceability

Review the checks for:

• Evidence IDs that do not exist
• Findings without evidence_ids
• Paths without an allowed path_type or Evidence reference
• Work items and timeline entries pointing to unknown Evidence
• Unlinked Evidence records
• Validated Findings with low confidence

An offline observation may use repro_command: n/a only when its notes field explicitly documents the offline limitation.

Use JSON when another tool needs stable fields:

python3 skills/case-review/scripts/review_case.py work/<case> --format json

建议下一步（选一个编号）

1. 补写缺失的 Evidence，并保留原始命令

2. 将候选 Finding 绑定到 Evidence 后重新审查

3. 为调用链或攻击链补充 P-id 和 Path 步骤

4. 生成 Markdown handoff summary

5. 换回 PRIMARY skill 继续分析

Phase 3: Fixity verification

When an Evidence record contains both content_hash and artifact_path, verify the case-local artifact:

python3 skills/case-review/scripts/review_case.py work/<case> --verify-hashes --strict

The script accepts sha256:<64 hex characters> and checks that the artifact remains inside the case root. A hash mismatch is a hard failure.

The PowerShell Evidence helper can record a hash while appending a record:

powershell -File skills/scripts/append-evidence.ps1 -CaseRoot work\<case> -Id E-001 -Title "Sample hash" -ReproCommand "sha256sum evidence/sample.bin" -ArtifactPath "evidence\sample.bin"

建议下一步（选一个编号）

1. 修复 hash mismatch 或替换已污染的工作副本

2. 为未固定的原始文件补充 SHA-256 和 artifact_path

3. 继续进入报告生成阶段

4. 导出 JSON 结果供 CI 保存

5. 暂停并请求人工复核

Phase 4: Handoff

Use strict mode before a final report or specialist handoff:

python3 skills/case-review/scripts/review_case.py work/<case> --strict --format markdown > work/<case>/report/case-review.md

The command is read-only with respect to the case unless shell redirection is explicitly used to save its output. The review is not legal advice and does not replace organizational evidence handling procedures.

建议下一步（选一个编号）

1. 将通过的 review 结果交给 docs-generator/ 生成正式报告

2. 回到 PRIMARY skill 补齐新的分析证据

3. 归档 Markdown 和 JSON review 结果

4. 暂停并请求人工复核

Language behavior contract

• Internal reasoning, tool selection, and phase control: English.
• User-visible messages, section labels, reports, and next-step menus: Chinese unless the user requests another language.
• Default bilingual labels place Chinese first and English second, separated by /.

Bootstrap boundary

This skill has no third-party dependency. If Python 3 is unavailable, the only allowed recovery action is the repository bootstrap path when a Python capability is registered for the current platform. If no such capability is registered, stop and report the missing runtime. Do not guess executable paths, download packages, or perform a manual install from inside this skill.

Routing context

Upstream entry: any reverse, forensics, CTF, or authorized security skill that has produced a case package.

Downstream exit: docs-generator/ for a formal report, or the original PRIMARY skill when the graph is incomplete.

Related modules: ops/evidence-finding-path.md, ops/timeline-workitem.md, digital-forensics/, reverse-engineering/, and docs-generator/.

References

• [NIST SP 800-86: Guide to Integrating Forensic Techniques into Incident Response](https://csrc.nist.gov/pubs/sp/800/86/final)
• [SWGDE Best Practices for Computer Forensic Acquisitions](https://www.swgde.org/documents/published-complete-listing/17-f-002-2-1/)
• [SWGDE Best Practices for Archiving Digital and Multimedia Evidence](https://www.swgde.org/documents/published-complete-listing/19-f-003-best-practices-for-archiving-digital-and-multimedia-evidence/)

任务完成自检

• [ ] 我是否审查了 scope.md、timeline.md、workitems.md 和 evidence/？
• [ ] 所有 Finding 是否引用了现存 Evidence？
• [ ] 所有 Path 是否包含合法 path_type 和 Evidence 引用？
• [ ] 是否执行了 hash verification，或记录了未执行原因？
• [ ] 是否以 strict 模式重新运行并保存了 review 结果？

Limitations

• Expects a structured case layout (scope, evidence, findings); ad-hoc notes need pre-organization.
• Reviews documentation quality, not the technical correctness of findings.

> Adapted from [zhaoxuya520/reverse-skill](https://github.com/zhaoxuya520/reverse-skill) (MIT).