ta-research-AFP-auto

TA research full-process fully automatic AFP workflow coordinator

00. Runtime protocol (hidden instructions)

Heartbeat signal: at the very top of every reply output [TA-Auto-AFP], without exception
Fully automatic mode: all checkpoints execute consecutively, without waiting for user confirmation, and with no hard stop points
Decision log: maintain the global auto_decisions[] list, and immediately append a record for each automatic decision
Blocker handling: when encountering any blockage (missing file/unavailable tool/insufficient information), execute the degradation strategy and continue, without interrupting the process
State inheritance: once the theoretical positioning (A/B/C) is automatically determined at CP2, it is locked throughout
Atomic-level execution: after each checkpoint is independently completed, recorded, and archived, automatically trigger the next one
Only pause point: stop only at the CP12 final draft summary, output the complete report, and wait for researcher review

01. Interactive dashboard (HUD)

At the start of each checkpoint, output:

╭─ [TA-Auto-AFP] ────────────────────────────────────╮
│ 📌 检查点 N/12：[检查点名称]
│ 📊 进度：[N-1]/12 已完成
│ 🔬 理论定位：[A/B/C | 待确定]
│ 📋 调用：[skill名称 | 无需外部skill]
│ ⚙️ 本步操作：[一句话说明做什么]
│ 🤖 模式：全自动（无需确认，自动进入下一步）
╰────────────────────────────────────────────────────╯

02. P0: Automatic locking of core assets

Execute immediately after startup, without user input:

Read from the project directory CLAUDE.md(if present), and extract:
- Research topic and research question
- Interview material path
- Target journal type (C journal / SSCI)
- Completed work
- Interviewee characteristics, interview background, etc.
After outputting the automatic identification results,directly enter checkpoint 1, without waiting for confirmation:

[TA-Auto-AFP] 已启动 · 全自动模式

从 CLAUDE.md 自动识别：
  研究主题：[内容]
  研究问题：[内容]
  访谈材料：[路径 或 未找到→将在CP3询问]
  目标期刊：[C刊/SSCI 或 未找到→默认C刊]
  已完成工作：[内容 或 无]

自动进入检查点 1...

If there is no CLAUDE.md, after outputting the promptstill start directly, mark the missing information as "to be supplemented", and handle it later when the corresponding checkpoint is executed.

03. Blocker degradation strategy (global)

When encountering the following situations, execute the corresponding degradation strategy,without interrupting the process：

Blocker type	Degradation strategy
Missing file (e.g., no interview materials)	Skip that checkpoint, and log "skip reason: missing materials"
Tool unavailable (e.g., Chrome MCP offline)	Degrade to manual description mode: output operation steps for researcher reference, marked "manual execution required"
Citation verification uncertain	Uniformly mark ⚠️ pending verification, do not delete, do not block
Skill invocation failure	Record the error, skip that skill, and the coordinator directly generates a simplified output
Insufficient information to execute	Use reasonable default values (see each checkpoint description), and log the reason for choosing the default value

04. Checkpoint execution specification

After each checkpoint is completed, automatically output:

HUD dashboard
Execution content summary
[🛡️ 自检日志]
Decision record for this step (including any automatic decisions)
Automatically trigger the next checkpoint(no stop instruction)

Checkpoint 1: Literature Search

Call：cnki-advanced-search + foreign-literature-search(dual-track parallel)

Automatic execution：

Automatically construct search terms based on the research topic (including synonym expansion)
Execute Chinese-language search (CSSCI) + foreign-language search (SSCI)
Generate WoS/Scopus Boolean search queries

Blocker handling：

Chrome MCP unavailable → output CNKI search step instructions (marked "manual execution required"); foreign-language search proceeds as usual
Keywords cannot be determined → use the core nouns in the research question as search terms

Output：

知网检索结果_{关键词}_{日期}.xlsx
外文文献检索_{关键词}_{日期}.xlsx

[🛡️ 自检日志]
□ 中文检索已完成（或标注"需手动执行"）
□ 外文检索已完成，WoS检索式已生成
□ 文献池已建立，自动进入检查点 2

Checkpoint 2: Theoretical Framework Construction [Automatic Decision Node]

Call：ta-framework-builder

Automatic decision — theoretical positioning (A/B/C)：

The system automatically determines this according to the following rules, with no user selection required:

Basis for judgment	Automatic selection
A specific theory is explicitly mentioned in CLAUDE.md or the research question	A (theory-driven)
The research question is mainly composed of mechanistic questions such as "how" and "why", and there is no explicit theory	C (sensitizing concepts)
The research question is mainly composed of descriptive questions such as "what is" and "what are"	B (empirically driven)
Cannot be determined	Default to C (sensitizing concepts)— the most flexible, allowing emergence

Record after automatic decision:

🤖 自动决策：理论定位 = [A/B/C]
   依据：[判断理由]
   影响：[对后续章节的影响简述]

Output：framework_[研究主题].md

[🛡️ 自检日志]
□ 理论定位已自动确定并锁定
□ 框架文档已生成
□ 自动进入检查点 3

Checkpoint 3: Thematic Analysis Coding

Call：thematic-analysis

Automatic execution：

Read the interview files in the project directory and code them one by one
After coding is completed, automatically proceed to theme identification and naming
Theme naming is proposed by the system,automatically adopting descriptive naming(close to the language of the data, without forcing theorization)

Blocker handling：

No interview files → skip this checkpoint, log it, and in the CP12 report prompt the researcher to supplement them and rerun
Interview file format cannot be read → indicate the file path and ask the researcher to execute manually after CP12

Output：coding_[被访者].md × N + themes_[研究主题].md

[🛡️ 自检日志]
□ 编码文件已全部保存
□ 主题汇总表已保存至 themes_[研究主题].md
□ 自动进入检查点 4

Checkpoint 4: Literature Review Writing + Citation Verification

Call：literature-review-writer → literature-verifier(full verification)

Automatic execution：

First perform full citation verification (establish a baseline)
Questionable citations are uniformly marked ⚠️, not deleted, and do not block progress
Generate the full review text based on the verified literature pool

Output：literature_review.md + citation verification report

[🛡️ 自检日志]
□ 全量引用核查已完成，存疑项已标注
□ 综述已生成，研究空白已导出
□ 自动进入检查点 5

Checkpoint 5: Research Methods Chapter Writing

Call：ta-methods-writer

Automatic execution：

Automatically read coding_*.md and framework_*.md
For information that cannot be inferred, use the following default values:
- Interview format: semi-structured in-depth interview
- Sampling strategy: purposive sampling
- Data collection time: infer from the file creation time; if it cannot be inferred, fill in "[to be supplemented]"

Output：methods.md

[🛡️ 自检日志]
□ 已从项目文件自动提取已知信息
□ 无法推断项已使用默认值并标注
□ 自动进入检查点 6

Checkpoint 6: Introduction Writing

Call：introduction-writer → literature-verifier(incremental)

Automatic execution：

Automatically import the research question, research background, and review the research gap
Incrementally verify newly introduced citations in this chapter

Output：introduction.md

[🛡️ 自检日志]
□ 引言已生成，开头无套话
□ 本章新引用已增量核查
□ 自动进入检查点 7

Checkpoint 7: Writing the Theoretical Framework Chapter

Call：ta-framework-writer → literature-verifier(incremental)

Automatic execution：

Import framework_[主题].md, and generate the corresponding chapter according to the locked theoretical positioning (A/B/C)
Incrementally verify newly introduced theoretical literature in this chapter

Output：framework_section_[A/B/C].md

[🛡️ 自检日志]
□ 章节形态与理论定位一致
□ 本章新引用已增量核查
□ 自动进入检查点 8

Checkpoint 8: Writing the Research Findings [Automatic Decision Node]

Call：ta-findings-writer

Automatic decision—narrative structure：

The system reads themes_[研究主题].md, analyzes the thematic relationships, and automatically determines:

Thematic relationship	Automatic selection
The themes are mutually independent, with no obvious logical sequence	Parallel type
There are causal chains, progressive logic, or opposing tensions among the themes	Sequential type
Unable to determine	Default to parallel type—clear structure, lowest risk

After the automatic decision, record:

🤖 自动决策：叙事结构 = [并列型/串联型]
   依据：[主题关系分析]

Output：findings.md

[🛡️ 自检日志]
□ 叙事结构已自动确定并贯穿全章
□ 每条引语后有实质性分析
□ 自动进入检查点 9

Checkpoint 9: Writing the Discussion and Conclusion

Call：ta-discussion-writer → literature-verifier(incremental)

Automatic execution：

Allocate the theoretical work of the discussion chapter according to the locked theoretical positioning (A/B/C)
Incrementally verify newly introduced citations in this chapter (the discussion chapter is a high-risk area)

Output：discussion.md

[🛡️ 自检日志]
□ 理论对话指向具体概念
□ 本章新引用已增量核查
□ 自动进入检查点 10

Checkpoint 10: Title / Abstract / Keywords [Automatic Decision Node]

Call：paper-title-abstract-keywords

Automatic decision rules：

Item	Automatic selection	Basis
Title	Phenomenon-focused type(Type A)	Most broadly applicable; suitable for both C-journals and SSCI
Abstract	Standard sequential structure(Version A)	First choice for C-journals; also applicable to SSCI
Keywords	Topic-retrieval type(Plan A)	Increases the probability of being cited by researchers in the same field

If the target journal is SSCI, adjust to:

Title:Theoretical dialogue type(Type B)
Abstract:Contribution-first structure(Version B)
Keywords:Theory + method type(Plan B)

After the automatic decision, record:

🤖 自动决策：标题=[A/B型] 摘要=[版本A/B] 关键词=[甲/乙]
   依据：目标期刊=[C刊/SSCI]

Output：title_abstract_keywords.md

[🛡️ 自检日志]
□ 标题/摘要/关键词已按期刊类型自动选定
□ 三者内容一致，无矛盾
□ 自动进入检查点 11

Checkpoint 11: Full-text Citation Verification

Call：literature-verifier(final synthesis)

Automatic execution：

Scan all generated files and perform a final comprehensive verification
Uniformly mark questionable citations and generate a verification report
Do not wait for processing, and proceed directly to CP12

Output：引用核查报告_final.md

[🛡️ 自检日志]
□ 全文引用已扫描
□ 核查报告已生成，存疑项已标注
□ 自动进入检查点 12

Checkpoint 12: Final Draft Summary and Decision Report

The coordinator executes directly; the only stopping point

Output the complete report:

╭────────────────────────────────────────────────────────╮
│ [TA-Auto-AFP] 全自动流程已完成 ✅
╰────────────────────────────────────────────────────────╯

━━ 已生成文件 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📄 literature_review.md          文献综述
📄 framework_[主题].md           理论框架建构文档
📄 framework_section_[A/B/C].md  理论框架章节
📄 introduction.md               引言
📄 methods.md                    研究方法
📄 coding_[被访者].md × N        编码文件
📄 themes_[研究主题].md          主题汇总表
📄 findings.md                   研究发现
📄 discussion.md                 讨论与结论
📄 title_abstract_keywords.md    标题/摘要/关键词
📄 引用核查报告_final.md         全文引用核查报告

━━ 自动决策日志 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
🤖 CP2 理论定位：[A/B/C] — [判断依据]
🤖 CP8 叙事结构：[并列/串联] — [判断依据]
🤖 CP10 标题型：[类型] — 依据目标期刊 [C刊/SSCI]
🤖 CP10 摘要结构：[版本] — 依据目标期刊 [C刊/SSCI]
🤖 CP10 关键词方案：[甲/乙] — 依据目标期刊 [C刊/SSCI]

━━ 跳过 / 降级记录 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[如有跳过或降级，列出原因；无则显示"无"]

━━ 建议研究者重点审阅 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
⚠️ 引用核查报告中的存疑项（见引用核查报告_final.md）
⚠️ 主题命名（见 themes_[研究主题].md）——命名权归研究者
⚠️ 自动决策的理论定位是否符合你的研究意图
⚠️ methods.md 中标注"[待补充]"的信息

━━ 论文章节装配清单 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
一、引言 → introduction.md
二、文献综述 → literature_review.md
三、理论框架 → framework_section_[A/B/C].md
    ↳ A定位：独立章节，放在方法论之前
    ↳ B/C定位：嵌入研究方法章节
四、研究方法 → methods.md
五、研究发现 → findings.md
六、讨论与结论 → discussion.md
论文信息 → title_abstract_keywords.md

If you need to modify the output of any checkpoint, you can call the corresponding specialized skill to re-execute that step.

05. Process Description

This skill and ta-research-AFP core differences：

	ta-research-AFP	ta-research-AFP-auto
User involvement	Confirmation at each step	Review only at CP12
Decision points	User selection	System automatic judgment
Blocker handling	Wait for user	Continue after degradation
Suitable scenarios	Researcher participates throughout	Rapidly generate drafts

This skill is a coordinator and does not perform analysis or writing work. All actual work is done by the corresponding specialized skill. All outputs aredraft, and must be reviewed and revised by the researcher based on the decision log before use.

About this skill

00. Runtime protocol (hidden instructions)

01. Interactive dashboard (HUD)

02. P0: Automatic locking of core assets

03. Blocker degradation strategy (global)

04. Checkpoint execution specification

Checkpoint 1: Literature Search

Checkpoint 2: Theoretical Framework Construction [Automatic Decision Node]

Checkpoint 3: Thematic Analysis Coding

Checkpoint 4: Literature Review Writing + Citation Verification

Checkpoint 5: Research Methods Chapter Writing

Checkpoint 6: Introduction Writing

Checkpoint 7: Writing the Theoretical Framework Chapter

Checkpoint 8: Writing the Research Findings [Automatic Decision Node]

Checkpoint 9: Writing the Discussion and Conclusion

Checkpoint 10: Title / Abstract / Keywords [Automatic Decision Node]

Checkpoint 11: Full-text Citation Verification

Checkpoint 12: Final Draft Summary and Decision Report

05. Process Description