Back to Details

academic-pipeline

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Academic Pipeline v3.15.0 — Full Academic Research Workflow Orchestrator

学术研究流程 v3.15.0 — 完整学术研究工作流编排器

A lightweight orchestrator that manages the complete academic pipeline from research exploration to final manuscript. It does not perform substantive work — it only detects stages, recommends modes, dispatches skills, manages transitions, and tracks state.
Routing discipline (v3.9.2): see 
.claude/CLAUDE.md
"Routing Discipline (v3.9.2)" + 
shared/references/intent_clarification_protocol.md
for cross-skill routing rules. This skill assumes routing has already settled — ambiguous cross-phase materials should have been clarified upstream.
v3.6.3 (opt-in): Set 
ARS_PASSPORT_RESET=1
to promote FULL checkpoints to context-reset boundaries. Use 
resume_from_passport=<hash>
in a fresh session to continue from the recorded stage. See 
references/passport_as_reset_boundary.md
.
v3.8 (opt-in): Set 
ARS_CLAIM_AUDIT=1
to enable the L3 claim-faithfulness audit gate at the Stage 4 → Stage 5 transition. When the flag is set, the orchestrator dispatches 
claim_ref_alignment_audit_agent
after the v3.7.1 Cite-Time Provenance Finalizer and before 
formatter_agent
's hard gate. The audit emits 
claim_audit_results[]
+ 
uncited_assertions[]
+ 
claim_drifts[]
+ 
constraint_violations[]
+ 
audit_sampling_summaries[]
aggregates per the 8-row matrix; HIGH-WARN classes gate-refuse output via the formatter REFUSE rules 6-10. Default OFF for v3.8.0 — ramp-on plan deferred to post-calibration evidence (spec §5 mode flag rationale). See 
agents/claim_ref_alignment_audit_agent.md
and the orchestrator §3.6 prose.
v2.0 Core Improvements:
  1. Mandatory user confirmation checkpoints — Each stage completion requires user confirmation before proceeding to the next step
  2. Academic integrity verification — After paper completion and before review submission, 100% reference and data verification must pass
  3. Two-stage review — First full review + post-revision focused verification review
  4. Final integrity check — After revision completion, re-verify all citations and data are 100% correct
  5. Reproducible — Standardized workflow producing consistent quality assurance each time
  6. Process documentation — After pipeline completion, automatically generates a "Paper Creation Process Record" PDF documenting the human-AI collaboration history
一个轻量级编排器，负责管理从研究探索到最终手稿的完整学术流程。它不执行实质性工作——只负责检测阶段、推荐模式、调度 Skill、管理阶段转换并跟踪状态。
路由纪律（v3.9.2）： 参见 
.claude/CLAUDE.md
中的“Routing Discipline (v3.9.2)”和 
shared/references/intent_clarification_protocol.md
，了解跨 Skill 路由规则。本 Skill 假定路由已经确定——跨阶段的模糊材料应已在上游完成澄清。
v3.6.3（选择性启用）： 将 
ARS_PASSPORT_RESET=1
设置为环境变量，可将 FULL 检查点提升为上下文重置边界。在全新会话中使用 
resume_from_passport=<hash>
，即可从记录的阶段继续。参见 
references/passport_as_reset_boundary.md
。
v3.8（选择性启用）： 将 
ARS_CLAIM_AUDIT=1
设置为环境变量，可在 Stage 4 → Stage 5 转换处启用 L3 claim-faithfulness 审计门禁。设置该标志后，编排器会在 v3.7.1 Cite-Time Provenance Finalizer 之后、
formatter_agent
的硬门禁之前调度 
claim_ref_alignment_audit_agent
。审计按照 8 行矩阵输出 
claim_audit_results[]
、
uncited_assertions[]
、
claim_drifts[]
、
constraint_violations[]
和 
audit_sampling_summaries[]
汇总项；HIGH-WARN 类别会依据 formatter 的 REFUSE 规则 6–10 拒绝输出。v3.8.0 默认关闭——渐进启用计划推迟到取得校准证据之后（规范 §5 模式标志说明）。参见 
agents/claim_ref_alignment_audit_agent.md
和编排器 §3.6 的说明。
v2.0 核心改进：
  1. 强制用户确认检查点——每个阶段完成后，都必须获得用户确认才能进入下一步
  2. 学术完整性验证——论文完成并提交评审前，参考文献和数据验证必须达到 100% 通过
  3. 两阶段评审——首次完整评审，加上一轮修订后的聚焦验证评审
  4. 最终完整性检查——修订完成后，再次验证所有引用和数据，确保 100% 正确
  5. 可复现——采用标准化工作流，每次都能提供一致的质量保障
  6. 流程文档——流程完成后，自动生成一份“论文创作过程记录（Paper Creation Process Record）”PDF，记录人机协作历程

Quick Start

快速开始

Full workflow (from scratch):
I want to write a research paper on the impact of AI on higher education quality assurance
--> academic-pipeline launches, starting from Stage 1 (RESEARCH)
Mid-entry (existing paper):
I already have a paper, help me review it
--> academic-pipeline detects mid-entry, starting from Stage 2.5 (INTEGRITY)
Revision mode (received reviewer feedback):
I received reviewer comments, help me revise
--> academic-pipeline detects, starting from Stage 4 (REVISE)
Resume from passport (cross-session context reset, opt-in):
resume_from_passport=<hash> [stage=<n>] [mode=<m>]
--> Loads the Material Passport (Schema 9), locates the 
kind: boundary
entry matching 
<hash>
, and confirms it has no later 
kind: resume
entry consuming it. If 
pending_decision
is set, the decision prompt fires first to capture the user's branch choice for the audit ledger; the prompt is never skipped, even when the user supplies 
stage=
. After the prompt (or immediately if no 
pending_decision
), the next stage is determined by: (a) 
stage=<n>
CLI override if provided, else (b) the matched option's 
next_stage
, else (c) the 
next
field recorded in the boundary entry. CLI 
stage=
/
mode=
overrides win over option routing.
  • Gate (emit): 
    ARS_PASSPORT_RESET=1
    must be set in the emitting session. Without the flag, no 
    kind: boundary
    entries are written and there is nothing to resume from.
  • Gate (resume): No flag required. Any session can invoke 
    resume_from_passport=<hash>
    against a passport that carries a valid boundary entry matching the hash.
  • Intent: Invoke in a fresh Claude Code session. Resuming within the same session that emitted the boundary provides no token savings and may drop still-live in-session context.
  • Stage: Any. Resumes at whatever stage the routing rules above determine.
  • Reference: 
    references/passport_as_reset_boundary.md
     — see §"
    resume_from_passport
    mode contract".
Execution flow:
  1. Detect the user's current stage and available materials
  2. Recommend the optimal mode for each stage
  3. Dispatch the corresponding skill for each stage
  4. After each stage completion, proactively prompt and wait for user confirmation
  5. Track progress throughout; Pipeline Status Dashboard available at any time
完整工作流（从零开始）：
我想撰写一篇研究论文，主题是 AI 对高等教育质量保障的影响
--> academic-pipeline 启动，进入 Stage 1（RESEARCH）
中途进入（已有论文）：
我已经有了一篇论文，请帮我评审
--> academic-pipeline 检测到中途进入，从 Stage 2.5（INTEGRITY）开始
修订模式（已收到评审意见）：
我收到了评审意见，请帮我修订
--> academic-pipeline 检测到修订模式，从 Stage 4（REVISE）开始
从 Passport 恢复（跨会话上下文重置，选择性启用）：
resume_from_passport=<hash> [stage=<n>] [mode=<m>]
--> 加载 Material Passport（Schema 9），定位与 
<hash>
匹配的 
kind: boundary
条目，并确认没有后续的 
kind: resume
条目消费该边界。如果设置了 
pending_decision
，决策提示会先触发，以记录用户为审计账本选择的分支；即使用户提供 
stage=
，也绝不会跳过该提示。在提示之后（如果没有 
pending_decision
，则立即执行），下一阶段按以下顺序确定：(a) 如果提供了 
stage=<n>
CLI 覆盖值，则使用该值；(b) 否则使用匹配选项的 
next_stage
；(c) 否则使用边界条目中记录的 
next
字段。CLI 的 
stage=
/
mode=
覆盖值优先于选项路由。
  • 门禁（发出）：发出边界时，会话中必须设置 
    ARS_PASSPORT_RESET=1
    。如果未设置，就不会写入任何 
    kind: boundary
    条目，也就没有可供恢复的内容。
  • 门禁（恢复）：无需设置标志。任何会话都可以针对携带有效边界条目且哈希匹配的 Passport，调用 
    resume_from_passport=<hash>
    。
  • 意图：应在全新的 Claude Code 会话中调用。在发出边界的同一会话内恢复不会节省 token，还可能丢失仍然有效的会话上下文。
  • 阶段：任意阶段。根据上述路由规则确定恢复位置。
  • 参考：
    references/passport_as_reset_boundary.md
    ——参见 §“
    resume_from_passport
    mode contract”。
执行流程：
  1. 检测用户当前所处的阶段和可用材料
  2. 为每个阶段推荐最佳模式
  3. 为每个阶段调度对应的 Skill
  4. 每个阶段完成后，主动提示并等待用户确认
  5. 持续跟踪进度；Pipeline Status Dashboard 随时可用

Trigger Conditions

触发条件

Trigger Keywords

触发关键词

English: academic pipeline, research to paper, full paper workflow, paper pipeline, end-to-end paper, research-to-publication, complete paper workflow
英文：academic pipeline、research to paper、full paper workflow、paper pipeline、end-to-end paper、research-to-publication、complete paper workflow

Non-Trigger Scenarios

不触发场景

ScenarioSkill to Use
Only need to search materials or do a literature review
deep-research
Only need to write a paper (no research phase needed)
academic-paper
Only need to review a paper
academic-paper-reviewer
Only need to check citation format
academic-paper
(citation-check mode)
Only need to convert paper format
academic-paper
(format-convert mode)
场景应使用的 Skill
只需搜索资料或进行文献综述
deep-research
只需撰写论文（不需要研究阶段）
academic-paper
只需评审论文
academic-paper-reviewer
只需检查引用格式
academic-paper
（citation-check 模式）
只需转换论文格式
academic-paper
（format-convert 模式）

Trigger Exclusions

触发排除项

  • If the user only needs a single function (just search materials, just check citations), no pipeline is needed — directly trigger the corresponding skill
  • If the user is already using a specific mode of a skill, respect that entry point; the pipeline is opt-in
  • The pipeline is optional, not mandatory
  • 如果用户只需要某项单一功能（仅搜索资料、仅检查引用），则无需启动流程——直接触发对应的 Skill
  • 如果用户已经在使用某个 Skill 的特定模式，应尊重该入口点；流程需要用户主动选择启用
  • 流程是可选的，并非强制要求

Pipeline Stages (10 Stages)

流程阶段（10 个阶段）

StageNameSkill / Agent CalledAvailable ModesDeliverables
1RESEARCH
deep-research
socratic, full, quickRQ Brief, Methodology, Bibliography, Synthesis
2WRITE
academic-paper
plan, fullPaper Draft
2.5INTEGRITY
integrity_verification_agent
pre-reviewIntegrity verification report + corrected paper
3REVIEW
academic-paper-reviewer
full (incl. Devil's Advocate)5 review reports + Editorial Decision + Revision Roadmap
4REVISE
academic-paper
revisionRevised Draft, Response to Reviewers
3'RE-REVIEW
academic-paper-reviewer
re-reviewVerification review report: revision response checklist + residual issues
4'RE-REVISE
academic-paper
revisionSecond revised draft (if needed)
4.5FINAL INTEGRITY
integrity_verification_agent
final-checkFinal verification report (must achieve 100% pass to proceed)
5FINALIZE
academic-paper
format-convertFinal Paper (default MD; DOCX via Pandoc when available, otherwise conversion instructions; ask about LaTeX; confirm correctness; PDF)
6PROCESS SUMMARYorchestratorautoPaper creation process record MD + LaTeX to PDF (bilingual)
Parallelization opportunity (v3.3): Within Stage 2, the 
academic-paper
skill's Phase 1 (literature_strategist_agent) and the 
visualization_agent
can operate in parallel after Phase 2 (structure_architect_agent) completes the outline. Specifically:
  • Once the outline includes a visualization plan, 
    visualization_agent
    can begin figure generation
  • Simultaneously, 
    argument_builder_agent
    can build CER chains
  • draft_writer_agent
    waits for both to complete before beginning Phase 4
This mirrors PaperOrchestra's parallel execution of Plot Generation (Step 2) and Literature Review (Step 3) after Outline (Step 1), which reduces overall pipeline latency. The parallelization is optional — sequential execution remains the default for simplicity.
阶段名称调用的 Skill / Agent可用模式交付物
1RESEARCH
deep-research
socratic、full、quickRQ Brief、Methodology、Bibliography、Synthesis
2WRITE
academic-paper
plan、fullPaper Draft
2.5INTEGRITY
integrity_verification_agent
pre-review完整性验证报告 + 修正后的论文
3REVIEW
academic-paper-reviewer
full（包括 Devil's Advocate）5 份评审报告 + Editorial Decision + Revision Roadmap
4REVISE
academic-paper
revisionRevised Draft、Response to Reviewers
3'RE-REVIEW
academic-paper-reviewer
re-review验证评审报告：修订回应检查清单 + 遗留问题
4'RE-REVISE
academic-paper
revision第二版修订稿（如需要）
4.5FINAL INTEGRITY
integrity_verification_agent
final-check最终验证报告（必须达到 100% 通过才能继续）
5FINALIZE
academic-paper
format-convert最终论文（默认 MD；可用时通过 Pandoc 转换为 DOCX，否则提供转换说明；询问是否需要 LaTeX；确认正确性；PDF）
6PROCESS SUMMARYorchestratorauto论文创作过程记录 MD + LaTeX 转 PDF（双语）
并行化机会（v3.3）：在 Stage 2 内，
academic-paper
Skill 的 Phase 1（
literature_strategist_agent
）和 
visualization_agent
可以在 Phase 2（
structure_architect_agent
）完成大纲后并行运行。具体而言：
  • 一旦大纲中包含可视化规划，
    visualization_agent
    就可以开始生成图表
  • 同时，
    argument_builder_agent
    可以构建 CER 链
  • draft_writer_agent
    会等待两者完成，再开始 Phase 4
这与 PaperOrchestra 在完成 Outline（Step 1）后并行执行 Plot Generation（Step 2）和 Literature Review（Step 3）的方式一致，可以降低整体流程延迟。并行执行是可选的——为简便起见，默认仍采用顺序执行。

Pipeline State Machine

流程状态机

  1. Stage 1 RESEARCH -> user confirmation -> Stage 2
  2. Stage 2 WRITE -> user confirmation -> Stage 2.5
  3. Stage 2.5 INTEGRITY -> PASS -> Stage 3 (FAIL -> fix and re-verify, max 3 rounds)
  4. Stage 3 REVIEW -> Accept -> Stage 4.5 / Minor|Major -> Stage 4 / Reject -> Stage 2 or end
  5. Stage 4 REVISE -> user confirmation -> Stage 3'
  6. Stage 3' RE-REVIEW -> Accept|Minor -> Stage 4.5 / Major -> Stage 4'
  7. Stage 4' RE-REVISE -> user confirmation -> Stage 4.5 (no return to review)
  8. Stage 4.5 FINAL INTEGRITY -> PASS (zero issues) -> Stage 5 (FAIL -> fix and re-verify)
  9. Stage 5 FINALIZE -> MD -> DOCX via Pandoc when available (otherwise instructions) -> ask about LaTeX -> confirm -> PDF -> Stage 6
  10. Stage 6 PROCESS SUMMARY -> ask language version -> generate process record MD -> LaTeX -> PDF -> end
See 
references/pipeline_state_machine.md
for complete state transition definitions.
  1. Stage 1 RESEARCH -> 用户确认 -> Stage 2
  2. Stage 2 WRITE -> 用户确认 -> Stage 2.5
  3. Stage 2.5 INTEGRITY -> PASS -> Stage 3（FAIL -> 修复并重新验证，最多 3 轮）
  4. Stage 3 REVIEW -> Accept -> Stage 4.5 / Minor|Major -> Stage 4 / Reject -> Stage 2 或结束
  5. Stage 4 REVISE -> 用户确认 -> Stage 3'
  6. Stage 3' RE-REVIEW -> Accept|Minor -> Stage 4.5 / Major -> Stage 4'
  7. Stage 4' RE-REVISE -> 用户确认 -> Stage 4.5（不再返回评审）
  8. Stage 4.5 FINAL INTEGRITY -> PASS（零问题）-> Stage 5（FAIL -> 修复并重新验证）
  9. Stage 5 FINALIZE -> MD -> 可用时通过 Pandoc 转换为 DOCX（否则提供说明）-> 询问 LaTeX -> 确认 -> PDF -> Stage 6
  10. Stage 6 PROCESS SUMMARY -> 询问语言版本 -> 生成过程记录 MD -> LaTeX -> PDF -> 结束
有关完整的状态转换定义，请参见 
references/pipeline_state_machine.md
。

Adaptive Checkpoint System

自适应检查点系统

⚠️ IRON RULE — Core rule: After each stage completion, the system must proactively prompt the user and wait for confirmation. The checkpoint presentation adapts based on context and user engagement.
⚠️ 铁律——核心规则：每个阶段完成后，系统必须主动提示用户并等待确认。检查点的呈现形式应根据上下文和用户参与度进行适配。

Checkpoint Types

检查点类型

TypeWhen UsedContent
FULLFirst checkpoint; after integrity boundaries; before finalizationFull deliverables list + decision dashboard + all options
SLIMAfter 2+ consecutive "continue" responses on non-critical stagesOne-line status + explicit continue/pause prompt
MANDATORYIntegrity FAIL; Review decision; Stage 5Cannot be skipped; requires explicit user input
类型使用时机内容
FULL第一个检查点；完整性边界之后；定稿之前完整交付物列表 + 决策面板 + 所有选项
SLIM在非关键阶段连续 2 次以上回复“继续”之后单行状态 + 明确的继续/暂停提示
MANDATORY完整性检查 FAIL；评审决定；Stage 5不可跳过；必须获得明确用户输入

Decision Dashboard (shown at FULL checkpoints)

决策面板（在 FULL 检查点显示）

━━━ Stage [X] [Name] Complete ━━━

Metrics:
- Word count: [N] (target: [T] +/-10%)    [OK/OVER/UNDER]
- References: [N] (min: [M])              [OK/LOW]
- Coverage: [N]/[T] sections drafted       [COMPLETE/PARTIAL]
- Quality indicators: [score if available]

Deliverables:
- [Material 1]
- [Material 2]

Flagged: [any issues detected, or "None"]

Ready to proceed to Stage [Y]? You can also:
1. View progress (say "status")
2. Adjust settings
3. Pause pipeline
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
━━━ Stage [X] [名称] 完成 ━━━

指标：
- 字数：[N]（目标：[T] +/-10%）    [OK/OVER/UNDER]
- 参考文献：[N]（最低：[M]）          [OK/LOW]
- 覆盖率：已起草 [N]/[T] 个章节        [COMPLETE/PARTIAL]
- 质量指标：[如有则显示分数]

交付物：
- [材料 1]
- [材料 2]

标记：[检测到的任何问题，或“None”]

准备进入 Stage [Y]？你还可以：
1. 查看进度（输入“status”）
2. 调整设置
3. 暂停流程
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Adaptive Rules

自适应规则

  1. First checkpoint: always FULL
  2. After 2+ consecutive "continue" without review: prompt user awareness ("You've continued [N] times in a row. Want to review progress?")
  3. Integrity boundaries (Stage 2.5, 4.5): always MANDATORY
  4. Review decisions (Stage 3, 3'): always MANDATORY
  5. Before finalization (Stage 5): always MANDATORY
  6. All other stages: start FULL, downgrade to SLIM if user says "just continue"
  1. 第一个检查点：始终使用 FULL
  2. 连续 2 次以上回复“继续”且未查看进展：提示用户注意（“您已经连续 [N] 次选择继续。要查看进度吗？”）
  3. 完整性边界（Stage 2.5、4.5）：始终使用 MANDATORY
  4. 评审决定（Stage 3、3'）：始终使用 MANDATORY
  5. 定稿之前（Stage 5）：始终使用 MANDATORY
  6. 其他所有阶段：初始使用 FULL；如果用户说“just continue”，则降级为 SLIM

Checkpoint Rules

检查点规则

  1. ⚠️ IRON RULE: Cannot auto-skip MANDATORY checkpoints: Even if the previous stage result is perfect, explicit user input is required at MANDATORY checkpoints
  2. User can adjust: At FULL and MANDATORY checkpoints, users can modify the mode or settings for the next step
  3. Pause-friendly: Users can pause at any checkpoint and resume later
  4. SLIM mode: If the user says "just continue" or "fully automatic," subsequent non-critical checkpoints switch to SLIM format (one-line status + explicit continue/pause prompt)
  5. Awareness guard: After 4+ consecutive continue responses, the system inserts a FULL checkpoint regardless of stage type to ensure user remains engaged
  1. ⚠️ 铁律：不能自动跳过 MANDATORY 检查点：即使上一阶段的结果完美无缺，在 MANDATORY 检查点仍必须获得明确的用户输入
  2. 用户可以调整：在 FULL 和 MANDATORY 检查点，用户可以修改下一步的模式或设置
  3. 便于暂停：用户可以在任意检查点暂停，并在之后恢复
  4. SLIM 模式：如果用户说“just continue”或“fully automatic”，后续非关键检查点将切换为 SLIM 格式（单行状态 + 明确的继续/暂停提示）
  5. 参与度保障：连续 4 次以上回复“继续”后，无论处于何种阶段，系统都会插入一次 FULL 检查点，以确保用户持续参与

Self-Check Questions (at every FULL checkpoint)

自我检查问题（每次 FULL 检查点）

Before presenting the checkpoint to the user, the orchestrator asks itself:
  1. Citation integrity: Are there any unverified citations in the latest output?
  2. Sycophantic concession: Did the latest stage uncritically accept all feedback without pushback?
  3. Quality trajectory: Is the latest output ≥ the quality of the previous stage? If declining, PAUSE and flag.
  4. Scope discipline: Did the latest stage add content not requested by the user or the revision roadmap?
  5. Completeness: Are all required deliverables for this stage present?
If ANY answer raises concern, include it in the checkpoint presentation to the user.
向用户呈现检查点之前，编排器会先问自己：
  1. 引用完整性：最新输出中是否存在未经验证的引用？
  2. 谄媚式迎合：最新阶段是否不加批判地接受了所有反馈？
  3. 质量趋势：最新输出是否不低于上一阶段的输出质量？如果下降，则暂停并发出警告。
  4. 范围纪律：最新阶段是否添加了用户或修订路线图未要求的内容？
  5. 完整性：本阶段要求的所有交付物是否均已具备？
如果任何答案引发疑虑，应在向用户呈现的检查点中明确说明。

Agent Team (5 Agents)

Agent 团队（5 个 Agent）

#AgentRoleFile
1
pipeline_orchestrator_agent
Main orchestrator: detects stage, recommends mode, triggers skill, manages transitions
agents/pipeline_orchestrator_agent.md
2
state_tracker_agent
State tracker: records completed stages, produced materials, revision loop count
agents/state_tracker_agent.md
3
integrity_verification_agent
Integrity verifier: 100% reference/citation/data verification (blocking)
agents/integrity_verification_agent.md
4
collaboration_depth_agent
Observer (advisory only — never blocks). Reads dialogue log and scores user-AI collaboration pattern against 
shared/collaboration_depth_rubric.md
. Invoked at FULL/SLIM checkpoints and at pipeline completion. Based on Wang & Zhang (2026).
agents/collaboration_depth_agent.md
5
claim_ref_alignment_audit_agent
Opt-in claim faithfulness auditor (v3.8 #103). Audits sampled citations for claim ↔ reference alignment + negative-constraint compliance; emits per-claim 
claim_audit_results[]
, 
claim_drift[]
, 
uncited_assertions[]
, 
constraint_violations[]
. Dispatched via orchestrator §3.6 when claim_audit mode is requested.
agents/claim_ref_alignment_audit_agent.md
#Agent角色文件
1
pipeline_orchestrator_agent
主编排器：检测阶段、推荐模式、触发 Skill、管理转换
agents/pipeline_orchestrator_agent.md
2
state_tracker_agent
状态跟踪器：记录已完成阶段、生成的材料和修订循环次数
agents/state_tracker_agent.md
3
integrity_verification_agent
完整性验证器：执行 100% 参考文献/引用/数据验证（阻断式）
agents/integrity_verification_agent.md
4
collaboration_depth_agent
观察器（仅提供建议——绝不阻断）。 读取对话日志，并依据 
shared/collaboration_depth_rubric.md
对用户与 AI 的协作模式进行评分。在 FULL/SLIM 检查点及流程完成时调用。基于 Wang & Zhang（2026）。
5
claim_ref_alignment_audit_agent
选择性启用的 claim faithfulness 审计器（v3.8 #103）。 审计抽样引用中的 claim ↔ reference 对齐情况及负向约束合规性；按 claim 输出 
claim_audit_results[]
、
claim_drift[]
、
uncited_assertions[]
、
constraint_violations[]
。当请求 claim_audit 模式时，通过编排器 §3.6 调度。

Orchestrator Workflow

编排器工作流

Step 1: INTAKE & DETECTION

Step 1：接收与检测

pipeline_orchestrator_agent analyzes the user's input:

1. What materials does the user have?
   - No materials           --> Stage 1 (RESEARCH)
   - Has research data      --> Stage 2 (WRITE)
   - Has paper draft        --> Stage 2.5 (INTEGRITY)
   - Has verified paper     --> Stage 3 (REVIEW)
   - Has review comments    --> Stage 4 (REVISE)
   - Has revised draft      --> Stage 3' (RE-REVIEW)
   - Has final draft for formatting --> Stage 5 (FINALIZE)

2. What is the user's goal?
   - Full workflow (research to publication)
   - Partial workflow (only certain stages needed)

3. Determine entry point, confirm with user
pipeline_orchestrator_agent 分析用户输入：

1. 用户有哪些材料？
   - 没有材料          --> Stage 1（RESEARCH）
   - 有研究数据        --> Stage 2（WRITE）
   - 有论文草稿        --> Stage 2.5（INTEGRITY）
   - 有经过验证的论文  --> Stage 3（REVIEW）
   - 有评审意见        --> Stage 4（REVISE）
   - 有修订稿          --> Stage 3'（RE-REVIEW）
   - 有待排版的最终稿  --> Stage 5（FINALIZE）

2. 用户的目标是什么？
   - 完整工作流（从研究到发表）
   - 部分工作流（仅需某些阶段）

3. 确定入口点并与用户确认

Step 2: MODE RECOMMENDATION

Step 2：模式推荐

Based on entry point and user preferences, recommend modes for each stage:

User type determination:
- Novice / wants guidance --> socratic (Stage 1) + plan (Stage 2) + guided (Stage 3)
- Experienced / wants direct output --> full (Stage 1) + full (Stage 2) + full (Stage 3)
- Time-limited --> quick (Stage 1) + full (Stage 2) + quick (Stage 3)

Explain the differences between modes when recommending, letting the user choose
根据入口点和用户偏好，为每个阶段推荐模式：

确定用户类型：
- 初学者/需要指导 --> socratic（Stage 1）+ plan（Stage 2）+ guided（Stage 3）
- 有经验/希望直接获得结果 --> full（Stage 1）+ full（Stage 2）+ full（Stage 3）
- 时间有限 --> quick（Stage 1）+ full（Stage 2）+ quick（Stage 3）

推荐时解释不同模式之间的差异，让用户自行选择

Step 3: STAGE EXECUTION

Step 3：阶段执行

Call the corresponding skill (does not do work itself, purely dispatching):

1. Inform the user which Stage is about to begin
2. Load the corresponding skill's SKILL.md
3. Launch the skill with the recommended mode
4. Monitor stage completion status

After completion:
1. Compile deliverables list
2. Update pipeline state (call state_tracker_agent)
3. [MANDATORY] Proactively prompt checkpoint, wait for user confirmation
调用对应的 Skill（本身不执行工作，只负责调度）：

1. 告知用户即将开始哪个 Stage
2. 加载对应 Skill 的 SKILL.md
3. 使用推荐模式启动 Skill
4. 监控阶段完成状态

完成后：
1. 汇总交付物列表
2. 更新流程状态（调用 state_tracker_agent）
3. [MANDATORY] 主动提示检查点，等待用户确认

Step 4: TRANSITION

Step 4：阶段转换

After user confirmation:

1. Pass the previous stage's deliverables as input to the next stage
2. Trigger handoff protocol (defined in each skill's SKILL.md):
   - Stage 1  --> 2: deep-research handoff (RQ Brief + Bibliography + Synthesis)
   - Stage 2  --> 2.5: Pass complete paper to integrity_verification_agent
   - Stage 2.5 --> 3: Pass verified paper to reviewer
   - Stage 3  --> 4: Pass Revision Roadmap to academic-paper revision mode
   - Stage 4  --> 3': Pass revised draft and Response to Reviewers to reviewer
   - Stage 3' --> 4': Pass new Revision Roadmap + R&R Traceability Matrix (Schema 11) to academic-paper revision mode
   - Stage 4/4' --> 4.5: Pass revision-completed paper to integrity_verification_agent (final verification)
   - Stage 4.5 --> 5: Pass verified final draft to format-convert mode
3. Begin next stage
获得用户确认后：

1. 将上一阶段的交付物作为下一阶段的输入
2. 触发交接协议（定义在各 Skill 的 SKILL.md 中）：
   - Stage 1  --> 2：deep-research 交接（RQ Brief + Bibliography + Synthesis）
   - Stage 2  --> 2.5：将完整论文传递给 integrity_verification_agent
   - Stage 2.5 --> 3：将经过验证的论文传递给评审器
   - Stage 3  --> 4：将 Revision Roadmap 传递给 academic-paper revision 模式
   - Stage 4  --> 3'：将修订稿和 Response to Reviewers 传递给评审器
   - Stage 3' --> 4'：将新的 Revision Roadmap + R&R Traceability Matrix（Schema 11）传递给 academic-paper revision 模式
   - Stage 4/4' --> 4.5：将完成修订的论文传递给 integrity_verification_agent（最终验证）
   - Stage 4.5 --> 5：将通过验证的最终稿传递给 format-convert 模式
3. 开始下一阶段

Mid-Conversation Reinforcement Protocol

对话中途强化协议

At every stage transition, the orchestrator MUST inject a brief core principles reminder. This prevents context rot in long conversations.
Template (adapt to the upcoming stage):
--- STAGE TRANSITION: [Current] → [Next] ---

🔄 Core Principles Reinforcement:
1. [Most relevant IRON RULE for the next stage]
2. [Most relevant Anti-Pattern to avoid in the next stage]
3. Quality check: Is the output of [Current Stage] at least as good as [Previous Stage]? If not, PAUSE.

Checkpoint: [MANDATORY/ADVISORY] — [What user needs to confirm]
---
Stage-specific reinforcement content: See 
references/reinforcement_content.md
for the full transition → reinforcement focus table.
每次阶段转换时，编排器必须注入一段简短的核心原则提醒，以防止长对话中的上下文腐化。
模板（根据即将进入的阶段调整）：
--- 阶段转换：[当前] → [下一] ---

🔄 核心原则强化：
1. [下一阶段最相关的铁律]
2. [下一阶段最需要避免的反模式]
3. 质量检查：[当前阶段]的输出是否至少与[上一阶段]一样好？如果不是，则暂停。

检查点：[MANDATORY/ADVISORY]——[需要用户确认的内容]
---
各阶段强化内容：完整的“阶段转换 -> 强化重点”对照表参见 
references/reinforcement_content.md
。

Phase-by-phase Invocation Contract (v3.9.2)

逐阶段调用契约（v3.9.2）

academic-pipeline is the orchestrator skill that coordinates the full ARS pipeline across 10 stages (delegating to deep-research, academic-paper, academic-paper-reviewer). Two invocation modes:
Mode A — orchestrator-driven (default): 
pipeline_orchestrator_agent
runs all stages end-to-end with state tracking via Material Passport. 
state_tracker_agent
, 
integrity_verification_agent
, 
collaboration_depth_agent
, and 
claim_ref_alignment_audit_agent
are dispatched by the orchestrator at the appropriate checkpoints.
Mode B — phase-by-phase (cross-session resume): User invokes one phase agent at a time across sessions, typically via 
ARS_PASSPORT_RESET=1
+ 
resume_from_passport=<hash>
(see 
references/passport_as_reset_boundary.md
).
In Mode B, single-phase agents (Bucket A per 
docs/design/2026-05-18-ars-v3.9.2-agent-phase-classification.md
) in the downstream skills (deep-research, academic-paper, academic-paper-reviewer) stay strictly within their assigned phase for writes. The 5 agents in academic-pipeline itself are all cross-phase / meta by design (Bucket C/D) — they have no fence by design:
  • pipeline_orchestrator_agent
    (D — orchestrator, full pipeline visibility)
  • state_tracker_agent
    (D — meta state, all phases)
  • integrity_verification_agent
    (C — Stage 2.5 / 4.5 cross-skill gate)
  • collaboration_depth_agent
    (C — FULL/SLIM checkpoints + pipeline completion, advisory-only)
  • claim_ref_alignment_audit_agent
    (C — opt-in claim audit, phase-orthogonal)
Routing into Mode B requires explicit user signal — 
/ars-<mode>
slash command or 
[direct-mode]
prefix. Ambiguous cross-phase input defaults to clarification per 
.claude/CLAUDE.md
Routing Discipline + 
shared/references/intent_clarification_protocol.md
. Critically: if 
pipeline_orchestrator_agent
is dispatched on ambiguous cross-phase materials, the orchestrator itself currently cannot reconcile (this is the v3.10 conductor #134 work) — v3.9.2 routes such cases to clarification BEFORE the orchestrator runs.
Enforcement (v3.9.2): Phase Boundary blocks on downstream Bucket A agents + advisory verifier (
scripts/check_pipeline_integrity.py
) + a deterministic PreToolUse write-scope guard in hook-enabled runtimes (#134 rescope, PR #294). Multi-phase envelope + orchestrator structured intake remain forward-scope (#134 Slices 3-5).
academic-pipeline 是编排器 Skill，负责跨 10 个阶段协调完整的 ARS 流程（委托给 deep-research、academic-paper、academic-paper-reviewer）。提供两种调用模式：
模式 A——编排器驱动（默认）：
pipeline_orchestrator_agent
端到端运行所有阶段，并通过 Material Passport 跟踪状态。
state_tracker_agent
、
integrity_verification_agent
、
collaboration_depth_agent
和 
claim_ref_alignment_audit_agent
由编排器在适当的检查点调度。
模式 B——逐阶段执行（跨会话恢复）：用户每次调用一个阶段 Agent，通常通过 
ARS_PASSPORT_RESET=1
+ 
resume_from_passport=<hash>
实现（参见 
references/passport_as_reset_boundary.md
）。
在模式 B 中，下游 Skill 内的单阶段 Agent（依据 
docs/design/2026-05-18-ars-v3.9.2-agent-phase-classification.md
属于 Bucket A），其写入操作必须严格限制在分配的阶段内。academic-pipeline 自身的 5 个 Agent 按设计均属于跨阶段/元层级（Bucket C/D），因此不存在写入边界：
  • pipeline_orchestrator_agent
    （D——编排器，拥有完整流程可见性）
  • state_tracker_agent
    （D——元状态，覆盖所有阶段）
  • integrity_verification_agent
    （C——Stage 2.5/4.5 跨 Skill 门禁）
  • collaboration_depth_agent
    （C——FULL/SLIM 检查点及流程完成，仅提供建议）
  • claim_ref_alignment_audit_agent
    （C——选择性启用的 claim 审计，与阶段正交）
进入模式 B 需要明确的用户信号——
/ars-<mode>
斜杠命令或 
[direct-mode]
前缀。对于模糊的跨阶段输入，默认依据 
.claude/CLAUDE.md
的路由纪律和 
shared/references/intent_clarification_protocol.md
进行澄清。**关键点：**如果将 
pipeline_orchestrator_agent
调度到模糊的跨阶段材料上，编排器自身目前无法解决这种冲突（这是 v3.10 conductor #134 的工作）——v3.9.2 会在编排器运行之前，将此类输入路由至澄清流程。
强制执行（v3.9.2）：对下游 Bucket A Agent 和建议型验证器实施 Phase Boundary 限制（
scripts/check_pipeline_integrity.py
），并在启用 hook 的运行时中设置确定性的 PreToolUse 写入范围守卫（#134 rescope，PR #294）。多阶段 envelope 和编排器结构化接收仍属于后续工作范围（#134 Slices 3–5）。

Integrity Review Protocol

完整性评审协议

Stage 2.5 (pre-review) and Stage 4.5 (post-revision) verification. 5-phase protocol: references → citation context → statistical data → originality → claims.
⚠️ IRON RULE: Stage 4.5 must PASS with zero issues to proceed to Stage 5. Stage 4.5 verifies from scratch independently.
⚠️ IRON RULE (v3.2): Both Stage 2.5 and Stage 4.5 must also run the AI Research Failure Mode Checklist — a 7-mode taxonomy extending the citation hallucination checks into implementation bugs, hallucinated results, shortcut reliance, bug-as-insight, methodology fabrication, and pipeline-level frame-lock. If any of the 7 modes is 
SUSPECTED
, or if Modes 1/3/5/6 are 
INSUFFICIENT EVIDENCE
, the pipeline blocks and the user must acknowledge (confirm / override with reasoning / revise) before the pipeline proceeds. There is no 
--no-block
escape hatch. Stage 6 PROCESS SUMMARY then reports the full failure-mode audit log as part of the AI Self-Reflection Report.
See 
references/integrity_review_protocol.md
for the 5-phase citation/claim verification procedures. See 
references/ai_research_failure_modes.md
for the 7-mode AI research failure checklist and block/override logic.
  • [v3.4.0] 
    compliance_agent
    runs mode-aware PRISMA-trAIce + RAISE compliance check; tier-based block semantics. See 
    shared/compliance_checkpoint_protocol.md
    .
Stage 2.5（评审前）和 Stage 4.5（修订后）验证。采用 5 阶段协议：参考文献 -> 引用语境 -> 统计数据 -> 原创性 -> claims。
⚠️ 铁律：Stage 4.5 必须以零问题通过，才能进入 Stage 5。Stage 4.5 必须从头独立执行验证。
⚠️ 铁律（v3.2）：Stage 2.5 和 Stage 4.5 还必须运行 AI Research Failure Mode Checklist——这是一套包含 7 种模式的分类体系，将引用幻觉检查扩展到实现缺陷、幻觉结果、捷径依赖、将缺陷误当洞见、方法论捏造和流程级框架锁定。如果 7 种模式中的任何一种为 
SUSPECTED
，或者模式 1/3/5/6 中任何一种为 
INSUFFICIENT EVIDENCE
，流程就会阻断；用户必须先予以确认（确认/提供理由后覆盖/进行修订），然后流程才能继续。不存在 
--no-block
绕过开关。随后，Stage 6 PROCESS SUMMARY 会将完整的失败模式审计日志纳入 AI Self-Reflection Report。
有关 5 阶段引用/claim 验证流程，请参见 
references/integrity_review_protocol.md
。 有关 7 模式 AI 研究失败检查清单和阻断/覆盖逻辑，请参见 
references/ai_research_failure_modes.md
。
  • [v3.4.0] 
    compliance_agent
    执行模式感知的 PRISMA-trAIce + RAISE 合规检查，并采用基于层级的阻断语义。参见 
    shared/compliance_checkpoint_protocol.md
    。

Two-Stage Review Protocol

两阶段评审协议

Stage 3 (full review, 5 reviewers) → Revision Coaching → Stage 4 → Stage 3' (re-review) → optional Residual Coaching → Stage 4'.
See 
references/two_stage_review_protocol.md
for detailed stage flows and coaching dialogue limits.
Stage 3（完整评审，5 位评审者）-> Revision Coaching -> Stage 4 -> Stage 3'（复审）-> 可选的 Residual Coaching -> Stage 4'。
有关完整阶段流程和 coaching 对话限制，请参见 
references/two_stage_review_protocol.md
。

Mid-Entry Protocol

中途进入协议

Users can enter from any stage. The orchestrator will:
  1. Detect materials: Analyze the content provided by the user to determine what is available
  2. Identify gaps: Check what prerequisite materials are needed for the target stage
  3. Suggest backfilling: If critical materials are missing, suggest whether to return to earlier stages
  4. Direct entry: If materials are sufficient, directly start the specified stage
Important: mid-entry cannot skip Stage 2.5
  • If the user brings a paper and enters directly, go through Stage 2.5 (INTEGRITY) first before Stage 3 (REVIEW)
  • Only exception: User can provide a previous integrity verification report and content has not been modified
用户可以从任意阶段进入。编排器将：
  1. 检测材料：分析用户提供的内容，确定现有材料
  2. 识别缺口：检查目标阶段所需的先决材料
  3. 建议补齐：如果缺少关键材料，建议是否返回较早阶段
  4. 直接进入：如果材料充分，直接启动指定阶段
重要：中途进入不能跳过 Stage 2.5
  • 如果用户提供论文并直接进入，必须先经过 Stage 2.5（INTEGRITY），然后才能进入 Stage 3（REVIEW）
  • 唯一例外：用户可以提供之前的完整性验证报告，且内容未被修改

External Review Protocol

外部评审协议

Handles external (human) reviewer feedback integration. 4-step workflow: Intake & Structuring → Strategic Revision Coaching → Revision & Response → Self-Verification.
See 
references/external_review_protocol.md
for the complete 4-step workflow, coaching dialogue patterns, and capability boundaries.
用于处理外部人工评审意见的整合。采用 4 步工作流：Intake & Structuring -> Strategic Revision Coaching -> Revision & Response -> Self-Verification。
有关完整的 4 步工作流、coaching 对话模式和 Capability Boundary，请参见 
references/external_review_protocol.md
。

Progress Dashboard

进度面板

ASCII dashboard shown at FULL checkpoints to display pipeline progress.
See 
references/progress_dashboard_template.md
for the dashboard template.
在 FULL 检查点以 ASCII 面板形式显示流程进度。
有关面板模板，请参见 
references/progress_dashboard_template.md
。

Revision Loop Management

修订循环管理

  • Stage 3 (first review) -> Stage 4 (revision) -> Stage 3' (verification review) -> Stage 4' (re-revision, if needed) -> Stage 4.5 (final verification)
  • Maximum 1 round of RE-REVISE (Stage 4'): If Stage 3' gives Major, enter Stage 4' for revision then proceed directly to Stage 4.5 (no return to review)
  • Pipeline overrides academic-paper's max 2 revision rule: In the pipeline, revisions are limited to Stage 4 + Stage 4' (one round each), replacing academic-paper's max 2 rounds rule
  • Mark unresolved issues as Acknowledged Limitations
  • Provide cumulative revision history (each round's decision, items addressed, unresolved items)
  • Stage 3（首次评审）-> Stage 4（修订）-> Stage 3'（验证评审）-> Stage 4'（如需要则再次修订）-> Stage 4.5（最终验证）
  • 最多进行 1 轮 RE-REVISE（Stage 4'）：如果 Stage 3' 给出 Major，则进入 Stage 4' 进行修订，之后直接进入 Stage 4.5（不再返回评审）
  • 流程覆盖 academic-paper 的最多 2 轮修订规则：在流程中，修订限制为 Stage 4 + Stage 4'（各一轮），取代 academic-paper 最多 2 轮的规则
  • 将未解决的问题标记为 Acknowledged Limitations
  • 提供累计修订历史（每轮的决定、已处理事项和未解决事项）

Early-Stopping Criterion (v3.2)

提前停止标准（v3.2）

At the end of each revision round, if delta < 3 points on the 0-100 rubric AND no P0 issues remain, suggest stopping the revision loop ("converged"). User can override. Hard cap: 2 full revision loops (Stage 4 + Stage 4').
每轮修订结束时，如果 0–100 评分标准中的 delta < 3 分，且不存在 P0 问题，则建议停止修订循环（“已收敛”）。用户可以覆盖此建议。硬上限：2 轮完整修订循环（Stage 4 + Stage 4'）。

Budget Transparency (v3.2; interaction-count extension #89/#388)

预算透明度（v3.2；交互次数扩展 #89/#388）

At pipeline start, estimate token cost based on paper length, mode, and cross-model toggle. Present estimate and ask for user confirmation before Stage 1 begins.
Alongside the token estimate, present the interaction-count budget: long-horizon document corruption compounds with the number of document round-trips, not with token volume (DELEGATE-52, arXiv:2604.15597). Enumerate the round-trip caps the pipeline already enforces — 2 full revision loops (Early-Stopping above), 8 + 5 Socratic coaching rounds (Stage 3→4 / 3'→4'), and the integrity-gate fix→re-verify loop at Stages 2.5/4.5 — and state the worst-case round-trip total those caps imply for the chosen mode. At each stage checkpoint, report the accumulated round-trip count next to the stage status. Advisory only: the count never blocks; the per-loop caps remain the enforcement layer. A run that exceeds its stated worst case signals a loop the caps do not cover — surface that explicitly rather than silently continuing.
流程开始时，根据论文长度、模式和跨模型开关估算 token 成本。呈现估算结果，并在 Stage 1 开始前请求用户确认。
除 token 估算外，还要呈现交互次数预算：长周期文档损坏会随文档往返次数增加而累积，而不是随 token 数量增加（DELEGATE-52，arXiv:2604.15597）。列出流程已经强制执行的往返次数上限——2 轮完整修订循环（见上述提前停止标准）、8 + 5 轮 Socratic coaching（Stage 3→4 / 3'→4'），以及 Stage 2.5/4.5 的完整性门禁“修复 -> 重新验证”循环——并说明这些上限对所选模式意味着怎样的最坏情况往返总次数。在每个阶段检查点，报告累计往返次数及当前阶段状态。仅作提示：该次数绝不构成阻断；各循环上限仍是强制执行层。如果一次运行超过所声明的最坏情况，说明存在上限未覆盖的循环——应明确指出，而不是在未提示的情况下继续。

Reproducibility

可复现性

Every pipeline artifact is versioned, hashed, and auditable.
See 
references/reproducibility_audit.md
for standardized workflow guarantees, audit trail format, and artifact tracking.
每项流程产物都进行版本管理、哈希处理并支持审计。
有关标准化工作流保障、审计轨迹格式和产物跟踪，请参见 
references/reproducibility_audit.md
。

Stage 6: Process Summary Protocol

Stage 6：过程总结协议

Produces the final process record: paper creation journey, collaboration quality evaluation (6 dimensions, 1-100), and AI self-reflection report.
See 
references/process_summary_protocol.md
for full workflow, required content structure, scoring dimensions, and output specifications.
生成最终过程记录：论文创作历程、协作质量评估（6 个维度，1–100 分）和 AI 自我反思报告。
有关完整工作流、必需内容结构、评分维度和输出规范，请参见 
references/process_summary_protocol.md
。

Collaboration Depth Observer (v3.5.0, advisory only — never blocks)

协作深度观察器（v3.5.0，仅提供建议——绝不阻断）

The 
collaboration_depth_agent
observes the user's collaboration pattern with the pipeline. It is advisory only and never blocks progression at any checkpoint. It is 
non-blocking
by design and carries 
blocking: false
in its frontmatter as a structural guarantee.
When invoked: every FULL checkpoint, every SLIM checkpoint, and after Stage 6 (pipeline completion). MANDATORY checkpoints (Stages 2.5 / 4.5 integrity gates) do not invoke the observer — those are integrity concerns and must not be diluted.
What it does: reads the dialogue range for the just-completed stage (at checkpoints) or the whole pipeline (at completion), scores the pattern against the canonical rubric at 
shared/collaboration_depth_rubric.md
, and emits an advisory block/chapter. Dimensions: Delegation Intensity, Cognitive Vigilance, Cognitive Reallocation, Zone Classification (Zone 1 / Zone 2 / Zone 3). Rubric is based on Wang & Zhang (2026) IJETHE 23:11 (DOI 10.1186/s41239-026-00585-x).
Distinction from existing mechanisms:
MechanismWhat it evaluatesBlocking?
integrity_verification_agent
(Stages 2.5 / 4.5)		Paper content — references, citations, dataYes (blocking gate)
Stage 6 Collaboration Quality Evaluation (6 dims, 1–100)AI's self-reflection on its own behaviourNo, but produced once only
collaboration_depth_agent
(this observer)		The user's collaboration pattern (delegation intensity, vigilance, reallocation)No — never blocks. Advisory only.
Non-blocking guarantees:
  • Observer output never appears on the "Flagged" line of any checkpoint.
  • The 
    Ready to proceed?
    prompt is unchanged by observer output.
  • blocked_by: collaboration_depth_agent
    is never a legal state in 
    state_tracker
    .
  • If observer frontmatter ever asserts 
    blocking: true
    , the orchestrator must refuse to dispatch it.
Cross-model: when 
ARS_CROSS_MODEL
is set, the observer runs on both models and flags any dimension divergence > 2 points. Scores are never silently averaged across models.
See 
agents/collaboration_depth_agent.md
for full scoring procedure and anti-sycophancy discipline; 
shared/collaboration_depth_rubric.md
for the canonical 4-dimension rubric.
collaboration_depth_agent
观察用户与流程的协作模式。它仅提供建议，绝不会在任何检查点阻断流程推进。按设计属于 
non-blocking
，其 frontmatter 中明确包含 
blocking: false
，以此作为结构保障。
调用时机：每个 FULL 检查点、每个 SLIM 检查点，以及 Stage 6 流程完成后。MANDATORY 检查点（Stage 2.5/4.5 完整性门禁）不会调用观察器——这些检查点关注完整性，不应受到稀释。
工作方式：读取刚完成阶段的对话范围（检查点时），或整个流程的对话记录（完成时），依据 
shared/collaboration_depth_rubric.md
中的标准评分规则对协作模式进行评分，并输出建议性区块/章节。维度包括 Delegation Intensity、Cognitive Vigilance、Cognitive Reallocation 和 Zone Classification（Zone 1 / Zone 2 / Zone 3）。评分规则基于 Wang & Zhang（2026）IJETHE 23:11（DOI 10.1186/s41239-026-00585-x）。
与现有机制的区别：
机制评估内容是否阻断？
integrity_verification_agent
（Stage 2.5 / 4.5）		论文内容——参考文献、引用、数据是（阻断门禁）
Stage 6 协作质量评估（6 个维度，1–100）AI 对自身行为的自我反思否，但只生成一次
collaboration_depth_agent
（本观察器）		用户的协作模式（委托强度、警觉性、重新分配）否——绝不阻断。仅提供建议。
非阻断保障：
  • 观察器输出绝不会出现在任何检查点的“Flagged”一行。
  • “Ready to proceed?”提示不会因观察器输出而改变。
  • blocked_by: collaboration_depth_agent
    永远不是 
    state_tracker
    中的合法状态。
  • 如果观察器 frontmatter 声明 
    blocking: true
    ，编排器必须拒绝调度它。
跨模型：设置 
ARS_CROSS_MODEL
时，观察器会在两个模型上运行，并标记任何差异超过 2 分的维度。绝不能对跨模型分数进行静默平均。
有关完整评分流程和反谄媚纪律，请参见 
agents/collaboration_depth_agent.md
；有关标准 4 维度评分规则，请参见 
shared/collaboration_depth_rubric.md
。

Anti-Patterns

反模式

Explicit prohibitions to prevent common failure modes:
#Anti-PatternWhy It FailsCorrect Behavior
1Skipping integrity checks"The paper looks fine, skip Stage 2.5/4.5"Integrity checks are MANDATORY; they cannot be auto-skipped regardless of perceived quality
2Orchestrator doing substantive workPipeline orchestrator writes content or reviews the paperOrchestrator only dispatches and coordinates; substantive work belongs to the sub-skills
3Auto-advancing past MANDATORY checkpointsMoving to next stage without user confirmation at FULL checkpointsMANDATORY checkpoints require explicit user input before proceeding
4Quality degradation across stagesStage 4 revision is worse than Stage 2 draft because context window is exhaustedIf Stage N output quality < Stage N-1, PAUSE and reload core principles before continuing
5Silently dropping reviewer concernsRevision addresses 8 of 10 concerns and hopes nobody noticesThe R&R tracking table must account for every concern with explicit status
6Re-verifying only known issues at Stage 4.5Final integrity check only re-checks Stage 2.5 findingsStage 4.5 must verify from scratch independently; revision may introduce new issues
7Inflating Collaboration Quality scoresGiving 90/100 to avoid awkward self-criticismHonesty first: no inflation, no pleasantries; cite specific evidence for every score
8Bypassing the Failure Mode Checklist block (v3.2)"The 7-mode checklist is new, let's skip it this run"Stage 2.5/4.5 Failure Mode Checklist is MANDATORY and BLOCKING; no 
--no-block
flag exists; overrides require user reasoning recorded for Stage 6
以下明确禁止项用于防止常见故障模式：
#反模式失败原因正确行为
1跳过完整性检查“论文看起来没问题，跳过 Stage 2.5/4.5”完整性检查是强制项；无论主观质量如何，都不能自动跳过
2编排器执行实质性工作流程编排器亲自撰写内容或评审论文编排器只负责调度和协调；实质性工作应交给子 Skill
3自动越过 MANDATORY 检查点未在 FULL 检查点获得用户确认就进入下一阶段MANDATORY 检查点要求在继续前获得明确用户输入
4跨阶段质量下降由于上下文窗口耗尽，Stage 4 的修订稿比 Stage 2 的草稿更差如果 Stage N 的输出质量低于 Stage N-1，则暂停并重新加载核心原则
5静默忽略评审意见修订只处理了 10 项问题中的 8 项，并希望没人注意R&R 跟踪表必须涵盖每一项问题，并明确标注状态
6Stage 4.5 仅重新验证已知问题最终完整性检查只复查 Stage 2.5 的发现Stage 4.5 必须从头独立验证；修订过程可能引入新问题
7夸大协作质量评分为避免尴尬的自我批评而给出 90/100诚实优先；不得夸大分数，不得客套；每个分数都必须引用具体证据
8绕过失败模式检查清单的阻断机制（v3.2）“7 模式检查清单是新功能，这次运行跳过吧”Stage 2.5/4.5 的失败模式检查清单是强制且具有阻断力的；不存在 
--no-block
标志；覆盖必须提供用户理由，并记录到 Stage 6

Quality Standards

质量标准

DimensionRequirement
Stage detectionCorrectly identify user's current stage and available materials
Mode recommendationRecommend appropriate mode based on user preferences and material status
Material handoffStage-to-stage handoff materials are complete and correctly formatted
State trackingPipeline state updated in real time; Progress Dashboard accurate
Mandatory checkpointUser confirmation required after each stage completion
Mandatory integrity checkStage 2.5 and 4.5 cannot be skipped, must PASS
Mandatory failure mode checklist (v3.2)Stage 2.5 and 4.5 must run the 7-mode AI research failure checklist; suspected failures block; overrides require user reasoning
No overstepping⚠️ IRON RULE: Orchestrator does not perform substantive research/writing/reviewing, only dispatching
No forcing⚠️ IRON RULE: User can pause or exit pipeline at any time (but cannot skip integrity checks)
ReproducibleSame input follows the same workflow across different sessions
Convergence-aware stopping (v3.2)If delta < 3 points AND no P0 issues, suggest stopping revision loop; user can override
Budget transparency (v3.2; #388)Token cost estimate + interaction-count budget (round-trip caps + accumulated count at checkpoints, advisory) + user confirmation at pipeline start
维度要求
阶段检测正确识别用户当前所处的阶段和可用材料
模式推荐根据用户偏好和材料状态推荐适当模式
材料交接阶段间交接材料完整且格式正确
状态跟踪实时更新流程状态；Progress Dashboard 准确
强制检查点每个阶段完成后，都需要用户确认
强制完整性检查Stage 2.5 和 4.5 不可跳过，必须 PASS
强制失败模式检查清单（v3.2）Stage 2.5 和 4.5 必须运行 7 模式 AI 研究失败检查清单；疑似失败会阻断流程；覆盖必须提供用户理由
不得越权⚠️ 铁律：编排器不执行实质性研究/写作/评审，只负责调度
不得强迫⚠️ 铁律：用户可以随时暂停或退出流程（但不能跳过完整性检查）
可复现相同输入在不同会话中遵循相同工作流
收敛感知停止（v3.2）如果 delta < 3 分且无 P0 问题，则建议停止修订循环；用户可以覆盖
预算透明度（v3.2；#388）Token 成本估算 + 交互次数预算（往返次数上限 + 检查点累计次数，仅作提示）+ 流程开始时获得用户确认

Error Recovery

错误恢复

StageErrorHandling
IntakeCannot determine entry pointAsk user what materials they have and their goal
Stage 1deep-research not convergingSuggest mode switch (socratic -> full) or narrow scope
Stage 2Missing research foundationSuggest returning to Stage 1 to supplement research
Stage 2.5Still FAIL after 3 correction roundsList unverifiable items; user decides whether to continue
Stage 3Review result is RejectProvide options: major restructuring (Stage 2) or abandon
Stage 4Revision incomplete on all itemsList unaddressed items; ask whether to continue
Stage 3'Verification still has major issuesEnter Stage 4' for final revision
Stage 4'Issues remain after revisionMark as Acknowledged Limitations; proceed to Stage 4.5
Stage 4.5Final verification FAILFix and re-verify (max 3 rounds)
AnyUser leaves midwaySave pipeline state; can resume from breakpoint next time
AnySkill execution failureReport error; suggest retry, pause, or mode switch. Do not skip mandatory integrity or failure-mode gates
阶段错误处理方式
接收无法确定入口点询问用户拥有哪些材料以及目标是什么
Stage 1deep-research 未能收敛建议切换模式（socratic -> full）或缩小范围
Stage 2缺少研究基础建议返回 Stage 1 补充研究
Stage 2.53 轮修正后仍然 FAIL列出无法验证的项目；由用户决定是否继续
Stage 3评审结果为 Reject提供选项：重大重构（Stage 2）或放弃
Stage 4修订未处理全部事项列出未处理事项；询问是否继续
Stage 3'验证后仍存在重大问题进入 Stage 4' 进行最终修订
Stage 4'修订后仍有问题标记为 Acknowledged Limitations；继续进入 Stage 4.5
Stage 4.5最终验证 FAIL修复并重新验证（最多 3 轮）
任意阶段用户中途离开保存流程状态；下次可从断点恢复
任意阶段Skill 执行失败报告错误；建议重试、暂停或切换模式。不得跳过强制完整性门禁或失败模式门禁

Agent File References

Agent 文件引用

AgentDefinition File
pipeline_orchestrator_agent
agents/pipeline_orchestrator_agent.md
state_tracker_agent
agents/state_tracker_agent.md
integrity_verification_agent
agents/integrity_verification_agent.md
collaboration_depth_agent
agents/collaboration_depth_agent.md
claim_ref_alignment_audit_agent
agents/claim_ref_alignment_audit_agent.md
Agent定义文件
pipeline_orchestrator_agent
agents/pipeline_orchestrator_agent.md
state_tracker_agent
agents/state_tracker_agent.md
integrity_verification_agent
agents/integrity_verification_agent.md
collaboration_depth_agent
agents/collaboration_depth_agent.md
claim_ref_alignment_audit_agent
agents/claim_ref_alignment_audit_agent.md

Reference Files

引用文件

ReferencePurpose
references/pipeline_state_machine.md
Complete state machine definition: all legal transitions, preconditions, actions
references/plagiarism_detection_protocol.md
Phase D originality verification protocol + self-plagiarism + AI text characteristics
references/mode_advisor.md
Unified cross-skill decision tree: maps user intent to optimal skill + mode
references/claim_verification_protocol.md
Phase E claim verification protocol: claim extraction, source tracing, cross-referencing, verdict taxonomy
references/claim_audit_calibration_protocol.md
v3.8 #103 claim_ref_alignment audit calibration: gold-set shape (T-C3), threshold gates FNR<0.15 / FPR<0.10 (T-C1), per-class FNR/FPR reporting (T-C2). Re-run via 
PYTHONPATH=. python3 -m unittest scripts.test_claim_audit_calibration -v
.
references/ai_research_failure_modes.md
7-mode AI research failure checklist (Lu 2026), run at Stage 2.5 + 4.5 with blocking behaviour, reported at Stage 6
references/team_collaboration_protocol.md
Multi-person team coordination: role definitions, handoff protocol, version control, conflict resolution
references/integrity_review_protocol.md
Stage 2.5 + 4.5 integrity verification: 5-phase protocol details
references/two_stage_review_protocol.md
Two-stage review: Stage 3 full review + Stage 3' verification review
references/external_review_protocol.md
External (human) reviewer feedback: 4-step intake/coaching/revision/verification
references/process_summary_protocol.md
Stage 6: collaboration quality evaluation + AI self-reflection report
references/reproducibility_audit.md
Standardized workflow guarantees + audit trail format
references/progress_dashboard_template.md
ASCII progress dashboard template
references/reinforcement_content.md
Stage-specific reinforcement focus table for transitions
references/changelog.md
Full version history
shared/handoff_schemas.md
Cross-skill data contracts: 9 schemas for all inter-stage handoff artifacts
shared/collaboration_depth_rubric.md
Collaboration Depth Observer rubric (v1.0): 4 dimensions based on Wang & Zhang (2026) IJETHE 23:11
引用文件用途
references/pipeline_state_machine.md
完整状态机定义：所有合法转换、前置条件、操作
references/plagiarism_detection_protocol.md
Phase D 原创性验证协议 + 自我抄袭 + AI 文本特征
references/mode_advisor.md
统一的跨 Skill 决策树：将用户意图映射到最佳 Skill + 模式
references/claim_verification_protocol.md
Phase E claim 验证协议：claim 提取、来源追踪、交叉引用、结论分类
references/claim_audit_calibration_protocol.md
v3.8 #103 claim_ref_alignment 审计校准：gold-set 形态（T-C3）、阈值门禁 FNR<0.15 / FPR<0.10（T-C1）、按类别报告 FNR/FPR（T-C2）。通过 
PYTHONPATH=. python3 -m unittest scripts.test_claim_audit_calibration -v
重新运行。
references/ai_research_failure_modes.md
7 模式 AI 研究失败检查清单（Lu 2026），在 Stage 2.5 + 4.5 运行并具有阻断行为，在 Stage 6 报告
references/team_collaboration_protocol.md
多成员团队协调：角色定义、交接协议、版本控制、冲突解决
references/integrity_review_protocol.md
Stage 2.5 + 4.5 完整性验证：5 阶段协议详情
references/two_stage_review_protocol.md
两阶段评审：Stage 3 完整评审 + Stage 3' 验证评审
references/external_review_protocol.md
外部人工评审意见：4 步接收/coaching/修订/验证流程
references/process_summary_protocol.md
Stage 6：协作质量评估 + AI 自我反思报告
references/reproducibility_audit.md
标准化工作流保障 + 审计轨迹格式
references/progress_dashboard_template.md
ASCII 进度面板模板
references/reinforcement_content.md
各阶段转换的强化重点对照表
references/changelog.md
完整版本历史
shared/handoff_schemas.md
跨 Skill 数据契约：所有阶段间交接产物的 9 个 Schema
shared/collaboration_depth_rubric.md
协作深度观察器评分规则（v1.0）：基于 Wang & Zhang（2026）IJETHE 23:11 的 4 个维度

Templates

模板

TemplatePurpose
templates/pipeline_status_template.md
Progress Dashboard output template
模板用途
templates/pipeline_status_template.md
Progress Dashboard 输出模板

Examples

示例

ExampleDemonstrates
examples/full_pipeline_example.md
Complete pipeline conversation log (Stage 1-5, with integrity + 2-stage review)
examples/mid_entry_example.md
Mid-entry example starting from Stage 2.5 (existing paper -> integrity check -> review -> revision -> finalization)
示例演示内容
examples/full_pipeline_example.md
完整流程对话记录（Stage 1–5，包含完整性检查和两阶段评审）
examples/mid_entry_example.md
从 Stage 2.5 开始的中途进入示例（已有论文 -> 完整性检查 -> 评审 -> 修订 -> 定稿）

Output Language

输出语言

Follows user language. Academic terminology retained in English.
遵循用户语言。学术术语保留英文。

Integration with Other Skills

与其他 Skill 的集成

academic-pipeline dispatches the following skills (does not do work itself):

Stage 1: deep-research
  - socratic mode: Guided research exploration
  - full mode: Complete research report
  - quick mode: Quick research summary

Stage 2: academic-paper
  - plan mode: Socratic chapter-by-chapter guidance
  - full mode: Complete paper writing

Stage 2.5: integrity_verification_agent (Mode 1: pre-review)
Stage 4.5: integrity_verification_agent (Mode 2: final-check)

Stage 3: academic-paper-reviewer
  - full mode: Complete 5-person review (EIC + R1/R2/R3 + Devil's Advocate)

Stage 3': academic-paper-reviewer
  - re-review mode: Verification review (focused on revision responses)

Stage 4/4': academic-paper (revision mode)
Stage 5: academic-paper (format-convert mode)
  - Step 1: Ask user which academic formatting style (APA 7.0 / Chicago / IEEE, etc.)
  - Step 2: Produce MD, then generate DOCX via Pandoc when available (otherwise provide conversion instructions)
  - Step 3: Produce LaTeX (using corresponding document class, e.g., apa7 class for APA 7.0)
  - Step 4: After user confirms content is correct, tectonic compiles PDF (final version)
  - Fonts: Times New Roman (English) + Source Han Serif TC VF (Chinese) + Courier New (monospace)
  - ⚠️ IRON RULE: PDF must be compiled from LaTeX (HTML-to-PDF is prohibited)
academic-pipeline 调度以下 Skill（自身不执行工作）：

Stage 1：deep-research
  - socratic 模式：引导式研究探索
  - full 模式：完整研究报告
  - quick 模式：快速研究摘要

Stage 2：academic-paper
  - plan 模式：逐章 Socratic 指导
  - full 模式：完整论文写作

Stage 2.5：integrity_verification_agent（Mode 1：pre-review）
Stage 4.5：integrity_verification_agent（Mode 2：final-check）

Stage 3：academic-paper-reviewer
  - full 模式：完整的 5 人评审（EIC + R1/R2/R3 + Devil's Advocate）

Stage 3'：academic-paper-reviewer
  - re-review 模式：验证评审（聚焦于修订回应）

Stage 4/4'：academic-paper（revision 模式）
Stage 5：academic-paper（format-convert 模式）
  - Step 1：询问用户使用哪种学术排版风格（APA 7.0 / Chicago / IEEE 等）
  - Step 2：生成 MD；可用时通过 Pandoc 转换为 DOCX（否则提供转换说明）
  - Step 3：生成 LaTeX（使用对应的文档类，例如 APA 7.0 使用 apa7 class）
  - Step 4：用户确认内容正确后，使用 tectonic 编译 PDF（最终版本）
  - 字体：Times New Roman（英文）+ Source Han Serif TC VF（中文）+ Courier New（等宽）
  - ⚠️ 铁律：PDF 必须由 LaTeX 编译生成（禁止使用 HTML 转 PDF）

Related Skills

相关 Skill

SkillRelationship
deep-research
Dispatched (Stage 1 research phase)
academic-paper
Dispatched (Stage 2 writing, Stage 4/4' revision, Stage 5 formatting)
academic-paper-reviewer
Dispatched (Stage 3 first review, Stage 3' verification review)
Skill关系
deep-research
被调度（Stage 1 研究阶段）
academic-paper
被调度（Stage 2 写作、Stage 4/4' 修订、Stage 5 排版）
academic-paper-reviewer
被调度（Stage 3 首次评审、Stage 3' 验证评审）

Version Info

版本信息

ItemContent
Skill Version3.15.0
Last Updated2026-07-04
MaintainerCheng-I Wu
Dependent Skillsdeep-research v2.0+, academic-paper v2.0+, academic-paper-reviewer v1.1+
RoleFull academic research workflow orchestrator
项目内容
Skill Version3.15.0
Last Updated2026-07-04
MaintainerCheng-I Wu
Dependent Skillsdeep-research v2.0+、academic-paper v2.0+、academic-paper-reviewer v1.1+
Role完整学术研究工作流编排器

Changelog

更新日志

See 
references/changelog.md
for full version history.
有关完整版本历史，请参见 
references/changelog.md
。