Menu
Peripheral vision audit — find what the author has become blind to in their results. Run in a FRESH TERMINAL.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
End-of-session shutdown sequence. Saves session log, updates MEMORY.md, updates CLAUDE.md, and stages a git commit.
First-time project setup. Checks prerequisites (Zotero MCP), asks for paper count, generates folder structure, and configures all templates.
Safely deprecate a file following the deprecation protocol.
Structured debate between two perspectives on a research design decision.
Find gaps in the literature based on existing Obsidian notes. Searches for high-quality articles the author may have missed.
Scaffold a new paper directory with the full structure, CLAUDE.md, and empty templates.
| name | blindspot |
| description | Peripheral vision audit — find what the author has become blind to in their results. Run in a FRESH TERMINAL. |
Based on Scott Cunningham's Blindspot skill (MixtapeTools). The Shklovsky Principle: by the time you've spent months on a paper, the main finding has collapsed your attention. Blindspot makes the stone stony again.
This skill MUST be run in a new Claude Code session with no prior context from the authoring sessions. The separation is what makes it work — the auditing Claude has no prior commitments to the author's narrative.
paper-N/paper/main.tex or the latest draftoutput/tables/, figures in output/figures/pre-analysis-plan/strategy-memo.mddata/codebook/variables.md, data/codebook/sample.md| Things Present in Output | Things Absent from Output | |
|---|---|---|
| Vices (problems) | Anomalies normalized — outliers not questioned, coefficient signs rationalized, patterns dismissed as noise | Checks never run — subgroups not split, placebo tests missing, robustness not attempted |
| Virtues (opportunities) | Richness ignored — heterogeneity patterns, mechanism evidence in appendix tables, interesting coefficients on controls | Extensions not considered — natural next questions the data can answer, adjacent literatures not connected |
For each finding, document:
File the report in paper-N/correspondence/blindspot/YYYY-MM-DD_blindspot-report.md
# Blindspot Report — Paper [N]
Date: YYYY-MM-DD
## Vices — Problems Hiding in Plain Sight
### [Finding 1]
- **Location:** Table 3, Column 4
- **Issue:** [Description]
- **Why missed:** [Author likely focused on...]
- **Action:** [Investigate / discuss / remove]
- **Severity:** Critical
## Vices — Checks Never Run
### [Finding 1]
...
## Virtues — Richness Being Ignored
### [Finding 1]
...
## Virtues — Extensions Not Considered
### [Finding 1]
...
## Summary
[N] findings: [X] critical, [Y] important, [Z] minor.