Coursework And Dissertation Support Design

Goal

Make Qiongli support undergraduate-and-above coursework and dissertation work as first-class academic project types while reusing the existing research lifecycle, writing harness, citation-risk policy, subject routing, literature workflows, and quality gates.

The new support should let users bring an assignment brief, marking rubric, learning outcomes, supervisor feedback, or dissertation handbook and receive a structured, academically honest workflow for planning, drafting, revising, and checking the work. Qiongli should stop treating every student-facing request as a journal manuscript. It should calibrate the output to the assignment, program level, word count, rubric, and institutional AI policy.

Current Context

The repository already has strong building blocks:

• /academic-write supports title, abstract, introduction, related work, literature review, research proposal, methodology, results, discussion, limitations, and conclusion drafting.
• proposal-writer explicitly covers thesis, dissertation, capstone, and course project proposals.
• /paper-write can build an outline, manuscript draft, claim-evidence map, and figures/tables plan from existing RESEARCH/[topic]/ artifacts.
• /lit-review, /paper-read, /find-gap, /study-design, /proofread, and /academic-present already cover much of a dissertation journey.
• The workflow contract already tracks stages, task IDs, canonical artifacts, stage handoff state, evidence ledgers, and claim-evidence integrity.
• The subject runtime now handles undergraduate-and-above business projects, but its quality target is still doctoral-level journal contribution.

The gap is product fit. Coursework and dissertations have different success criteria from manuscripts:

• The rubric matters more than journal novelty.
• Learning outcomes and module requirements can be binding constraints.
• Reflection and positionality may be valid evidence in some assignments.
• Word count, formatting, and submission requirements are strict.
• Students need support that is transparent about AI policy and academic integrity.
• Dissertations need chapter and supervisor-feedback state, not only paper sections.

Product Decision

Add two first-class workflows:

/coursework [assignment brief, task, or topic]
/dissertation [topic, program, or level]

These workflows should be product wrappers over the existing academic research engine rather than separate writing systems. They introduce a learning-aware front end and new artifact contracts, then route into existing research, literature, design, writing, proofread, and presentation workflows where those tools are a good fit.

The design uses two new supplemental stages:

• L: Coursework and learning-assessment support
• M: Dissertation and major-project support

L and M are not replacements for stages A-K. They are project-mode stages that can orchestrate existing tasks. A dissertation may use M1 for planning, then still run A1, B1, C1, F2, J4, and K1 as needed.

Scope

In scope:

• Add /coursework and /dissertation workflow documents.
• Add L and M stage definitions to the workflow contract source and generated contract documentation.
• Add coursework and dissertation artifacts under the existing RESEARCH/[topic]/ project root.
• Add coursework-specific templates for assignment brief intake, rubric maps, learning outcome maps, outlines, claim-evidence plans, draft structure, revision plans, and submission checklists.
• Add dissertation-specific templates for dissertation plans, chapter maps, chapter status, supervisor feedback logs, milestone plans, and defense prep.
• Add skill specs for assignment analysis, rubric mapping, coursework architecture, coursework revision, dissertation planning, chapter architecture, supervisor feedback integration, and dissertation readiness.
• Update platform routing so natural language requests for coursework, assignment briefs, module rubrics, dissertations, theses, capstones, and supervisor feedback route to the new workflows.
• Add academic-integrity and AI-policy checks that are visible in output.
• Add fixtures and tests for routing, artifact contracts, registry validation, and policy guardrails.

Out of scope for the first implementation:

• Do not build a grade predictor.
• Do not promise marks, grades, acceptance, or supervisor approval.
• Do not bypass university AI-use policies or produce hidden authorship.
• Do not invent citations, source quotations, page numbers, data, fieldwork, interview material, lab results, supervisor comments, or institutional rules.
• Do not make problem-set, exam, quiz, or timed-assessment solving the primary product surface. Qiongli can explain concepts and help structure reasoning, but the first release focuses on scholarly written coursework.
• Do not replace the existing /paper router.
• Do not turn every course case into a business subject activation. Coursework classification and subject classification remain separate decisions.

User-Facing Model

Coursework

Users can start from any of these:

/coursework Analyze this assignment brief.
/coursework Write a plan for a 2,000-word undergraduate essay.
/coursework Revise this case analysis against the rubric.
/coursework Build a presentation script from this module assignment.
/coursework Check whether this literature review meets the learning outcomes.

The workflow classifies the assignment into one of these coursework types:

• essay
• report
• case_analysis
• reflective_writing
• literature_review
• research_proposal
• presentation
• portfolio
• lab_or_methods_report
• capstone_project
• dissertation_component

The classification controls structure, evidence expectations, tone, and the kind of upstream research workflows that may be used.

Dissertation

Users can start from any of these:

/dissertation Plan an undergraduate dissertation on platform governance.
/dissertation Turn this proposal into a chapter plan.
/dissertation Revise Chapter 2 using supervisor feedback.
/dissertation Build a final readiness checklist for this dissertation.
/dissertation Prepare viva questions from my dissertation draft.

The workflow records the level:

• undergraduate
• taught_master
• professional_master
• research_master
• doctoral

The level controls contribution expectations. Undergraduate and taught master's work should emphasize feasibility, correct method use, literature grounding, and rubric fit. Doctoral work can reuse the stricter research and journal quality gates where appropriate.

New Task IDs

Stage L: Coursework And Learning Assessment

Task ID	Purpose	Primary output
L1	Assignment brief intake	assignment/brief.md
L2	Rubric and learning-outcome mapping	assignment/rubric_map.md, assignment/learning_outcomes.md
L3	Coursework outline and structure plan	coursework/outline.md
L4	Coursework claim-evidence and citation plan	coursework/claim_evidence_plan.md
L5	Coursework draft or section draft	coursework/draft.md
L6	Coursework revision against rubric	coursework/revision_plan.md
L7	Coursework final readiness check	assignment/submission_checklist.md

Stage M: Dissertation And Major Project

Task ID	Purpose	Primary output
M1	Dissertation project planning	dissertation/dissertation_plan.md
M2	Dissertation chapter architecture	dissertation/chapter_map.md
M3	Dissertation chapter drafting	dissertation/chapters/
M4	Supervisor feedback integration	dissertation/supervisor_feedback_log.md, dissertation/revision_plan.md
M5	Dissertation milestone and risk planning	dissertation/milestone_plan.md
M6	Dissertation final readiness check	dissertation/final_readiness.md
M7	Viva or defense preparation	dissertation/defense_prep.md

Artifact Model

Keep using RESEARCH/[topic]/ so existing context packaging, path conventions, and handoff tooling continue to work.

RESEARCH/[topic]/
├── assignment/
│   ├── brief.md
│   ├── rubric_map.md
│   ├── learning_outcomes.md
│   ├── academic_integrity_notes.md
│   └── submission_checklist.md
├── coursework/
│   ├── outline.md
│   ├── claim_evidence_plan.md
│   ├── citation_plan.md
│   ├── draft.md
│   ├── revision_plan.md
│   └── final_response.md
├── dissertation/
│   ├── dissertation_plan.md
│   ├── chapter_map.md
│   ├── chapter_status.md
│   ├── milestone_plan.md
│   ├── supervisor_feedback_log.md
│   ├── revision_plan.md
│   ├── final_readiness.md
│   ├── defense_prep.md
│   └── chapters/
└── context/
    ├── research_state.md
    ├── decision_log.md
    ├── boundary_review.md
    └── stage_handoff.md

The assignment/ directory is shared by coursework and dissertation projects because dissertations often have handbooks, module rules, marking criteria, and institutional AI-policy language.

Coursework Workflow Semantics

/coursework runs in preview-first mode.

Step 1: Intake

Parse or request:

• assignment title and module/program context,
• task prompt or brief,
• rubric or marking criteria,
• learning outcomes,
• word count and formatting constraints,
• citation style,
• allowed source types,
• deadline or milestone constraints,
• AI-use policy if available,
• user's current material.

If the brief or rubric is missing, continue with explicit missing-input flags. Do not invent module rules.

Step 2: Classify The Coursework

Classify the task type and level. The classifier should separate:

• coursework type,
• subject or method routing,
• project level,
• evidence expectations,
• whether personal reflection is an acceptable evidence source.

Example: a "business case assignment" is coursework type case_analysis. It does not automatically activate the business subject unless the request also contains scholarly business research signals.

Step 3: Map Rubric And Learning Outcomes

Create a rubric map that turns each criterion into:

• required capability,
• evidence or content needed,
• quality threshold,
• section where it should appear,
• current status,
• risk if missing.

For learning outcomes, record whether each outcome is addressed directly, indirectly, or not yet addressed.

Step 4: Build The Structure

Build an outline appropriate to the task type. Do not default to manuscript IMRaD structure unless the assignment asks for a research report.

Structure examples:

• Essay: thesis, conceptual setup, argument clusters, counterargument, conclusion.
• Report: executive summary, issue, analysis, recommendations, implementation or limitations.
• Case analysis: case facts, diagnostic frame, alternatives, recommendation, evidence limits.
• Reflection: experience anchor, analytical lens, learning outcome connection, implication, boundary.
• Literature review: search scope, themes, synthesis, gap, implication.

Step 5: Draft Under The Writing Harness

Use the existing writing harness:

• state the section's role,
• state the central claim and evidence threshold,
• draft in paragraph clusters,
• review for drift, support, specificity, and rubric fit,
• ask for missing user material when required.

Personal reflection, placement experience, or fieldwork claims require user-supplied facts. Qiongli must not invent lived experience.

Step 6: Revise Against The Rubric

The revision pass compares the draft to:

• rubric map,
• learning outcomes,
• word count,
• citation plan,
• academic integrity notes,
• assignment format.

The output should list concrete revisions and blocked items. It should not claim the work will receive a grade.

Step 7: Final Readiness

The final checklist confirms:

• every rubric criterion is addressed or flagged,
• every learning outcome is addressed or flagged,
• citations are present where claims need support,
• personal or empirical claims are user-supplied,
• word count and formatting constraints are visible,
• AI-use and disclosure requirements are recorded,
• missing information remains visible.

Dissertation Workflow Semantics

/dissertation is a wrapper around a long-running project state machine.

Step 1: Project Profile

Record:

• title or working topic,
• degree level,
• discipline or subject,
• department or program,
• dissertation type,
• word count and chapter expectations,
• handbook or marking criteria,
• supervisor requirements,
• ethics constraints,
• current stage,
• available materials.

Supported dissertation types:

• empirical quantitative,
• empirical qualitative,
• mixed methods,
• systematic or scoping review,
• theoretical or conceptual,
• design science or artifact-based,
• professional or applied project.

Step 2: Plan The Dissertation

M1 creates dissertation/dissertation_plan.md with:

• working title,
• research problem,
• research questions,
• expected contribution at the correct degree level,
• method or review approach,
• required approvals,
• source and data dependencies,
• timeline,
• risks and fallback options.

This may call existing tasks such as A1, A2, A4, C1, C3, C4, and D1 when the project needs stronger research framing or ethics planning.

Step 3: Build Chapter Architecture

M2 creates dissertation/chapter_map.md with:

• chapter purpose,
• target word count,
• inputs required,
• output expectations,
• dependencies on other chapters,
• evidence threshold,
• status.

Default chapter families:

• introduction,
• literature review,
• methodology,
• findings or results,
• discussion,
• conclusion,
• references and appendices.

The chapter map should adapt for systematic reviews, conceptual dissertations, and professional projects.

Step 4: Draft Chapters

M3 routes chapters through existing writing capabilities:

• literature review chapters can use /lit-review, /paper-read, and /academic-write,
• methodology chapters can use /study-design and /academic-write,
• results chapters can use analysis interpretation and tables/figures plans,
• discussion and conclusion can use the writing harness and claim-evidence map.

Every chapter draft must record unresolved evidence gaps and claim strength.

Step 5: Integrate Supervisor Feedback

M4 parses feedback into:

• issue,
• affected chapter or claim,
• required action,
• evidence needed,
• priority,
• status,
• response or revision note.

The feedback workflow must preserve the user's supervisor's meaning. It should not invent feedback or imply supervisor approval.

Step 6: Track Milestones And Risks

M5 tracks:

• proposal approval,
• ethics application,
• data collection,
• analysis,
• chapter draft deadlines,
• supervisor review windows,
• final formatting,
• submission.

Risks should include data access, recruitment, ethics, source availability, method fit, analysis complexity, and word count pressure.

Step 7: Final Readiness And Defense

M6 checks:

• chapter completeness,
• research question alignment,
• method and evidence fit,
• citation coverage,
• claim-evidence traceability,
• formatting and submission requirements,
• AI-use notes,
• unresolved risks.

M7 prepares viva or defense material when relevant:

• likely questions,
• contribution explanation,
• methods defense,
• limitations defense,
• what changed from proposal to final submission,
• concise oral summary.

Academic Integrity Contract

Add assignment/academic_integrity_notes.md for both workflows.

The contract should record:

• institution or course AI policy when supplied,
• disclosure requirement when supplied,
• what assistance Qiongli provided,
• which facts, experiences, data, or interpretations came from the user,
• missing policy information,
• blocked requests.

Mandatory behavior:

• Do not fabricate citations, quotations, page numbers, datasets, interviews, fieldnotes, lab results, placement experiences, supervisor feedback, ethics approval, or institutional rules.
• Do not hide uncertainty about sources or evidence.
• Do not promise grades.
• Do not present planned work as completed findings.
• Do not rewrite personal reflection as if Qiongli had the user's lived experience.
• When AI-policy information is missing, mark it as missing and advise the user to check the relevant course or institution policy before submission.

Allowed behavior:

• Explain the assignment.
• Build an outline.
• Map the rubric.
• Suggest sources to verify.
• Draft or revise sections from user-provided facts and evidence.
• Improve clarity, structure, argument, and citation placement.
• Produce checklists and feedback reports.

Skill And Template Additions

New skill specs

Add focused skills instead of expanding one large coursework file:

content/skills/L_coursework/assignment-brief-analyzer.md
content/skills/L_coursework/rubric-mapper.md
content/skills/L_coursework/coursework-architect.md
content/skills/L_coursework/coursework-reviser.md
content/skills/M_dissertation/dissertation-planner.md
content/skills/M_dissertation/chapter-architect.md
content/skills/M_dissertation/supervisor-feedback-integrator.md
content/skills/M_dissertation/dissertation-readiness-checker.md

Each skill should define inputs, outputs, constraints, failure modes, and domain-aware behavior. The skills should be added to content/skills/registry.yaml and generated documentation.

New templates

content/templates/assignment-brief.md
content/templates/rubric-map.md
content/templates/learning-outcomes.md
content/templates/academic-integrity-notes.md
content/templates/coursework-outline.md
content/templates/coursework-claim-evidence-plan.md
content/templates/coursework-revision-plan.md
content/templates/coursework-submission-checklist.md
content/templates/dissertation-plan.md
content/templates/dissertation-chapter-map.md
content/templates/dissertation-chapter-status.md
content/templates/supervisor-feedback-log.md
content/templates/dissertation-milestone-plan.md
content/templates/dissertation-final-readiness.md
content/templates/dissertation-defense-prep.md

Templates should contain explicit missing-information fields rather than blank prompts that encourage invented content.

Routing And Trigger Rules

Update platform routing so these requests route to /coursework:

• "assignment brief"
• "coursework"
• "module assignment"
• "marking rubric"
• "learning outcomes"
• "write my essay for this course" with policy guardrails
• "case analysis assignment"
• "reflective assignment"
• "portfolio assessment"
• "capstone coursework"

Route to /dissertation:

• "dissertation"
• "thesis"
• "undergraduate dissertation"
• "master's dissertation"
• "supervisor feedback"
• "chapter plan"
• "viva prep"
• "defense prep"
• "dissertation handbook"

Routing must preserve subject refinement:

• Coursework type decides the project workflow.
• Subject routing decides whether economics, finance, accounting, business, or another subject overlay should load.
• A business case assignment is not enough to activate business unless the scholarly business subject signals pass their normal gate.

Workflow Contract Changes

Update the canonical workflow contract source:

• Add stages L and M.
• Add task IDs L1-L7 and M1-M7.
• Add expected output artifacts.
• Add refresh points for coursework and dissertation context.
• Add academic_project_type or equivalent project metadata.

Suggested project metadata:

academic_project_type:
  - journal_manuscript
  - research_paper
  - coursework
  - capstone
  - dissertation
  - thesis
  - presentation

Do not replace existing paper_type. paper_type remains about research design shape. academic_project_type records the educational or publication context.

Context Continuity

Extend context state so long coursework and dissertation projects do not lose constraints:

• current assignment brief,
• rubric criteria,
• learning outcomes,
• AI policy status,
• degree level,
• chapter status,
• supervisor feedback,
• deadline and milestone constraints,
• user-supplied personal or empirical material,
• unresolved evidence gaps.

Stage handoff should include whether the next task is allowed to draft prose or must first request missing material.

Evaluation Strategy

Add fixtures that prove both positive routing and guardrails.

Coursework positives

• Undergraduate essay with rubric and word count routes to /coursework.
• Case analysis assignment routes to coursework, not automatically to business subject activation.
• Literature review coursework can call literature support without claiming systematic-review-grade coverage.
• Reflective assignment requires user-supplied experience before drafting personal claims.
• Research proposal coursework routes to coursework and can reuse proposal-writer.

Dissertation positives

• Undergraduate dissertation request routes to /dissertation.
• Master's dissertation chapter planning creates M1/M2 artifacts.
• Supervisor feedback request routes to M4.
• Dissertation viva prep routes to M7.
• Dissertation literature review can reuse /lit-review and /academic-write.

Guardrails

• Missing rubric produces gap notes rather than invented criteria.
• Missing AI policy creates an academic-integrity warning.
• Requests for guaranteed grades are blocked.
• Requests to fabricate citations are blocked.
• Personal reflection without user material is blocked or converted into questions for the user.
• Timed exam or quiz requests do not route to coursework drafting.

Expected Code And Content Changes

Workflow files

Add:

content/workflow/workflows/coursework.md
content/workflow/workflows/dissertation.md
content/workflow/references/stage-L-coursework.md
content/workflow/references/stage-M-dissertation.md

Update:

content/workflow/workflows/academic-write.md
content/workflow/workflows/paper.md
content/workflow/references/workflow-contract.md
content/workflow/references/coverage-matrix.md
content/workflow/references/platform-routing.md
content/standards/research-workflow-contract.yaml

Skills and registry

Add L_coursework and M_dissertation skill directories, then update:

content/skills/registry.yaml
content/skills-summary.md
content/skills-core.md
docs/reference/skills.md
docs/zh/reference/skills.md

Update schema enums if the skill schema validates stage names.

CLI and MCP surfaces

If the CLI exposes workflow names directly, add:

qiongli coursework
qiongli dissertation

If MCP exposes workflow routing through qiongli_task_run, add the two workflows to accepted task names and preview output.

The first implementation can keep these as skill/workflow routes without new provider tools.

Documentation

Update:

README.md
docs/guide/task-recipes.md
docs/guide/using-agent-skills.md
docs/zh/guide/task-recipes.md
docs/zh/guide/using-agent-skills.md
CLAUDE.md

Mention that coursework support is academic-assistance support with visible policy checks, not a grade guarantee.

Implementation Slices

Slice 1: Contracts and docs

• Add stages, task IDs, workflow docs, templates, and routing references.
• Add registry entries for new skills.
• Add generated docs if generation scripts are part of the normal workflow.
• Verify documentation and schema validation.

Slice 2: Coursework vertical path

• Implement /coursework with L1-L7.
• Add assignment brief analyzer, rubric mapper, coursework architect, and coursework reviser.
• Add coursework routing fixtures and guardrail tests.
• Verify a complete essay/report/case-analysis preview without external literature providers.

Slice 3: Dissertation vertical path

• Implement /dissertation with M1-M7.
• Add dissertation planner, chapter architect, supervisor feedback integrator, and dissertation readiness checker.
• Reuse existing research workflows for literature, study design, chapter drafting, proofread, and presentation prep.
• Verify an undergraduate dissertation plan and supervisor-feedback revision path.

Slice 4: Runtime and release hardening

• Add CLI/MCP task exposure where needed.
• Add cross-platform route checks.
• Add Chinese docs.
• Run release-readiness validation before packaging.

Verification

Required checks after implementation:

uv run python tooling/scripts/validate_research_standard.py
uv run python -m pytest

Focused checks should also cover:

• workflow contract generation,
• skill registry loading,
• platform routing,
• subject routing near-misses,
• coursework guardrails,
• dissertation artifact paths,
• docs links.

If the full test suite is too broad during implementation, each slice must define a smaller focused command and finish with the full validation before release.

Risks And Mitigations

Risk	Mitigation
Coursework support becomes generic ghostwriting	Require rubric, evidence, user material, and academic-integrity notes; block fabrication and grade guarantees
Rubric handling invents module rules	Keep missing rubric fields explicit and avoid inferred institutional policy
Subject routing over-activates from course wording	Keep coursework type classification separate from subject classification
New stages break schema validation	Update schema enums, registry docs, and contract generation in the same slice
Dissertation duplicates /paper	Make /dissertation a wrapper over existing A-K tasks, not a parallel research engine
First implementation is too large	Deliver in slices with contract/docs first, then coursework, then dissertation

Release Criteria

The feature is release-ready when:

• /coursework and /dissertation are documented workflow entry points.
• Stage L and M task IDs appear in the canonical workflow contract.
• Core templates exist and contain missing-information fields.
• Coursework and dissertation skills are in the registry.
• Routing fixtures prove positive and guardrail behavior.
• Guardrail tests block fabrication, grade guarantees, missing reflection material, and timed-assessment drafting.
• Existing /paper, /academic-write, and subject routing behavior still passes.
• Documentation explains the academic-integrity boundary clearly.