research trained
The learned layer of the research guru: question, primary source, observed claim, explicit inference, decision.
mental model
Question → primary source → observed claim → explicit inference → decision. Repomix can provide code context at the source stage; it cannot validate runtime facts.
examples
repomix --include "src/**,docs/**" --style markdown -o context.md
Then inspect the generated summary and content before the file crosses a trust boundary.
best practices
- Open primary sources for changing facts and link them beside the claim.
- Define the decision criteria before collecting options.
- Narrow repository packs and review output for sensitive material; Repomix documents git-aware filtering and security checks (source).
strengths
The process keeps conclusions falsifiable and makes sources reviewable. Repomix creates a compact input for a scoped code review.
weaknesses / pain points
Documentation can be incomplete or stale, and a snapshot cannot establish deployed or authenticated behavior. Security scanners reduce risk but cannot prove a pack contains no sensitive data.
gotchas
Search-result snippets are not evidence. A source's publication date and a runtime observation date answer different questions.
known bugs
No version-specific defect is known here.
troubleshooting
| symptom | root cause | fix |
|---|---|---|
| recommendation cannot be checked | source is missing or indirect | open and cite the primary page |
| context file exposes irrelevant content | pack scope is broad | use explicit includes and inspect output |
| source conflicts | criteria were implicit | state decision criteria and report the conflict |