| name | report-review |
| description | Review a SOLAR LINE episode or summary report for quality, accuracy, and readability. Checks data integrity, Japanese text quality, and orbital mechanics correctness. |
| argument-hint | ["episode number or \"summary\" or \"all\" or \"final\""] |
Report Quality Review
Two-stage systematic review for SOLAR LINE reports.
How to Use
Specify what to review with $ARGUMENTS:
1 through 5 โ draft review of a specific episode reportsummary โ draft review of all summary pagesall โ draft review of everythingfinal โ final review of deployed site (isolated, no codebase access)
Stage 1: Draft Review (with codebase access)
Reviewer Persona
IMPORTANT: All reviews must be conducted by agents with separate context (Task agents or Codex), NOT the current session. Set this system prompt for the reviewer:
ใใชใใฏSFไฝๅใจ่ป้ๅๅญฆใซ้ขๅฟใใใ่ชญ่
ใงใใใใฉใกใใซใๅฐ้็ใช็ฅ่ญใฏใใใพใใใ
ใSOLAR LINEใใจใใใขใใกใซใคใใฆไบๅ็ฅ่ญใฏไธๅใใใพใใใ
ใฌใใผใใๅใใฆ่ชญใไธ่ฌ่ชญ่
ใจใใฆใไปฅไธใฎ่ฆณ็นใงใฌใใฅใผใใฆใใ ใใ๏ผ
- ๅฐ้็จ่ชใฎ่ชฌๆใฏๅๅใ๏ผๅ่ฆใฎ่ชญ่
ใ็่งฃใงใใใ๏ผ
- ๅณ่กจใใใฃใผใใฏๅๆใใตใใผใใใฆใใใ๏ผไฝใ็คบใใฆใใใๆ็ขบใ๏ผ
- ๅๆใฎ่ซ็ๅฑ้ใฏ่ฟฝใใใ๏ผ้ฃ่บใฏใชใใ๏ผ
- ๅๆ็ฅ่ญใชใใงๅ
ๅฎนใๆฅฝใใใใ๏ผ
็็ดใซใใฃใผใใใใฏใใฆใใ ใใใ
Implementation
Use a Task agent (Sonnet) for the draft review:
Task(subagent_type="general-purpose", model="sonnet", prompt=`
[Insert reviewer persona above as system context]
Review the report at reports/data/episodes/ep0X.md (or summary/*.md).
Read the report data, check against the checklist below, and provide
specific, actionable feedback.
`)
For physics-specific review, use Codex (gpt-5.2 reasoning model):
/nice-friend Review the physics calculations in EP0X.
Are the orbital mechanics claims well-supported?
Are the ฮV, transfer time, and mass calculations consistent?
Automated Checks (run first)
cd ts && npm test
Manual Review Checklist
1. Data Integrity
2. Japanese Text Quality
3. Orbital Mechanics Accuracy
4. Report Structure
5. Narrative Plausibility
Stage 2: Final Review (deployed site, no codebase)
Purpose: Evaluate the reader experience of the deployed site without developer context.
Implementation
Launch a Task agent in isolation (no file access), using only WebFetch:
Task(subagent_type="general-purpose", model="sonnet", prompt=`
[Insert reviewer persona above]
You are reviewing the deployed website at:
https://sksat.github.io/solar-line/
Using WebFetch, visit the following pages and evaluate the reader experience:
1. Top page (index.html) โ is the project purpose clear?
2. Episode 1 report โ can you follow the analysis as a newcomer?
3. Cross-episode summary โ does it provide useful synthesis?
4. Navigation โ can you find what you need? Are links working?
Focus on:
- Layout and readability on the rendered page
- Navigation flow between pages
- Broken links or missing content
- Whether the analysis is accessible to the target audience
- Overall impression as a first-time visitor
Report your findings with specific page URLs and quotes.
`)
Final Review Checklist
Efficiency Notes
- Run automated tests first โ they catch most data integrity issues
- Use Sonnet subagents for draft review (separate context required)
- Use gpt-5.2 (reasoning model) for physics consultations via nice-friend
- Focus final review on reader experience, not data accuracy
- Don't fix issues during review โ create separate tasks for fixes
- Record review findings in ideas/ for future sessions
