Menu
Peripheral vision audit — find what the author has become blind to in their results. Run in a FRESH TERMINAL.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
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.