Research Workspace
Research is a folder per PID, not another CSV. It is the place for manual SERP review, AI-assisted analysis, competitor notes, saved sources, and the decision record. The CSV remains the index; the folder holds the evidence.
Create the folder
scs research --workspace bs --pid FS-0001
scs research --workspace bs --pid FS-0001 --check
scs research --workspace bs --pid FS-0001 --check --require-export
--require-export also fails when exports/ is empty. Use it when the brief
depends on a tool export.
scs research requires the PID to exist in <workspace>-keywords.csv.
Layout
data/<workspace>/research/<pid>/
├── README.md # question, hypotheses, decision, prompt log
├── serp.md # top results, H2 matrix, gaps, differentiation
├── competitors/ # one note per competitor URL
├── sources/ # saved pages, PDFs, citation notes
├── exports/ # raw tool exports for this PID (NeuronWriter, SpyFu, GKP, WriterZen)
├── assets/ # images and diagrams for the article
└── prompts/ # optional AI prompt and session logs
What --check requires
README.mdandserp.mdexist.README.mdhas## Questionand## Decision.serp.mdhas## Top results,## H2 matrix,## Gaps, and## Differentiation.serp.mdcontains at least 3 competitor URLs.- No
TODO,TBD,[insert,[fill, orlorem ipsumplaceholders.
The research check is a minimum evidence gate, not a quality score. The human decision still comes from the operator.
Working with opencode
Point opencode at the PID folder and ask it to fill serp.md from pasted SERP
data, extract the competitor H2 matrix, or draft candidate angles. Do not let it
invent URLs, statistics, or quotes. Record the prompts that materially shaped
the research in README.md or prompts/.
Git and backups
Keep README.md, serp.md, and competitors/*.md in Git. Keep large binaries
in sources/, exports/, and assets/ out of Git or use Git LFS. The VPS
backup should include the whole workspace.
A research folder is not a substitute for checking the live SERP. The
serp.md URLs must be real results you reviewed, not AI-generated examples.