| name | storage-literature-synthesizer |
| description | Read storage, database, and distributed-systems papers and turn them into reusable design constraints, assumptions, mechanisms, and positioning. Use when surveying a topic, extracting implementation ideas, comparing systems, or building related-work and novelty framing. |
Storage Literature Synthesizer
Overview
Use this skill to read papers like an implementer and a paper writer at the same time.
The output should be a design map, not a generic summary.
Workflow
- Read with a concrete question.
- Extract boundary, mechanism, and claim separately.
- Distinguish what is authoritative from what is optimization or packaging.
- Build a comparison table against adjacent systems.
- Translate the paper into reusable research pressure for your own work.
- Compress the result into reusable constraints, comparison axes, and related-work framing.
Reading Questions
Read papers by asking:
- What exact problem does this system narrow?
- What must be true for the design to work?
- Where is the truth boundary?
- Which parts are protocol, and which are engineering?
- What does the evaluation really prove?
- What would fail if one assumption changed?
Extraction Rules
For every paper, extract these fields:
- system goal
- state and ownership model
- core mechanism
- safety condition
- steady-state cost
- failure or degraded-mode story
- explicit non-goals
- strongest supported claim
- weakest part of the evidence
Use references/paper-digest-template.md for the exact template.
Positioning Rules
Do not position by adjectives such as "simpler", "better", or "more practical".
Position by axis:
- truth placement
- recovery authority
- compaction or indexing contract
- read/write amplification tradeoff
- metadata/control-plane split
- answerability or freshness semantics
Output Modes
Common outputs:
- one-paper design digest
- side-by-side comparison table
- design lineage from older systems to newer ones
- related-work section material
- novelty-boundary note for your own project
- paper cluster note for a concrete topic
Failure Modes To Avoid
- summarizing implementation detail without isolating the protocol
- repeating the abstract instead of identifying the real constraint
- claiming novelty where the paper only changes packaging
- copying benchmark numbers without checking the compared path or workload
Reference Loading
Load only what you need:
references/paper-digest-template.md for structured paper digestionreferences/system-landscape.md for industrial/open-source reference systems worth comparing againstreferences/venue-map.md for top conferences and journals by subfieldreferences/hotspot-reading.md for turning recent accepted papers and CFPs into a scoped reading plan
