| name | dissertation-learning-loop |
| description | Use when dissertation work should convert new reading, source searches, LMS/module materials, supervisor resources, or discussion outcomes into durable project knowledge, literature-map updates, proposal implications, and next-reading questions without enabling unsupervised background automation. |
Dissertation Learning Loop
Use this skill when the agent needs to learn from newly read sources, websites, LMS content, supervisor or public resource material, supervisor feedback, proposal discussions, or user-confirmed decisions and carry that learning into later dissertation work.
Purpose
Make the agent's learning task-triggered, source-grounded, and reusable. This is a local learning workflow, not autonomous background browsing.
Core Rule
Every learned item must keep its evidence boundary:
CONFIRMED: directly supported by a local file, LMS/module source, user-confirmed fact, or verified source.
LITERATURE-SUPPORTED: supported by a screened academic source.
CONTEXTUAL: useful for thinking, but not formal dissertation evidence.
INFERENCE: the agent's synthesis or implication.
TO CONFIRM: needs user, supervisor, LMS, or source verification.
When To Use
Use for:
- literature search follow-up
- reading-list or source-note ingestion
- proposal/gap/methodology discussion where new decisions emerge
- supervisor feedback digestion
- updating Obsidian, source registers, or research-wiki after new evidence
- preparing the next research-search questions
Do not use for:
- raw participant data storage
- inventing citation details from memory
- automatic weekly searches unless a separate automation is explicitly created
- writing formal text before source-first and document-quality gates are applied
Workflow
- Identify the learning input:
- source file, URL, LMS page, user-confirmed statement, supervisor note, or discussion result.
- Classify the learning type:
- literature concept
- project decision
- methodology implication
- ethics/design boundary
- writing-style preference
- next-reading need
- Extract only reusable knowledge:
- key concept or claim
- why it matters for this dissertation
- what it can support
- what it cannot support
- next question it raises
- Update the right layer:
knowledge-base/SOURCE_REGISTER.md for source status
knowledge-base/sources/ for source notes
research-wiki/LITERATURE_MAP.md for clusters and gap logic
research-wiki/METHOD_DECISIONS.md for methodology decisions
research-wiki/OPEN_QUESTIONS.md for unresolved questions
- Obsidian for navigation and thinking links
- Record major learning events in
research-wiki/TASK_STATE.md when they affect future work.
Output Shape
Learning captured:
- Source/input:
- Evidence status:
- Reusable point:
- Dissertation implication:
- Boundary:
- Files updated:
- Next reading/question:
Guardrails
- Do not treat a skimmed or metadata-only source as full evidence.
- Do not store full copyrighted text or raw participant material in project knowledge layers.
- Do not let a useful idea become a formal claim without citation readiness.
- If the learning changes the dissertation route, mark the affected proposal/methodology decision clearly.