Claude skills for seeing what others miss — read skeptically, trace consequences, anticipate failure.
Most Claude skills help you produce things. These help you not get fooled — by a document, by your own optimism, by the obvious-but-wrong answer.
Built for analysts, investors, founders, and PMs: people whose job is to look at a situation and see further into it than everyone else in the room.
Read between the lines of any document — surface what's not being said.
Hedging, deflection, conspicuous omissions, buried leads, revealing word frequencies. Built for earnings calls, job postings, press releases, and any text where the subtext beats the summary.
Ships with scripts/scan.py — an evidence collector that quantifies emphasis (what gets repeated), hedging density, and passive-voice signals to ground the read in numbers.
"What are they really saying?" · "Is this a red flag?" · "Decode this earnings call"
Reason past the obvious consequence.
Maps the cascading effects of a decision or event — second, third, and long-tail — and finds the non-obvious winners and losers everyone else missed.
"What are the knock-on effects?" · "Play this forward" · "Who actually benefits?"
Find what kills your plan — before you commit.
Assumes your project already failed, reasons backwards to the causes, ranks them by likelihood × impact, and prescribes the one fix that matters most. Based on Gary Klein's technique.
"Poke holes in this" · "What could go wrong?" · "Stress-test my plan"
your-skills-directory/
├── forensic-read/
│ ├── SKILL.md
│ └── scripts/
├── second-order/
│ └── SKILL.md
└── pre-mortem/
└── SKILL.md
git clone https://github.com/3243dwon/clear-eye.git
cp -r clear-eye/forensic-read clear-eye/second-order clear-eye/pre-mortem ~/.claude/skills/Install all three, or just the one you want — they're fully independent.
These are plain SKILL.md files plus optional Python scripts. They work across Claude.ai, Claude Code, and the Claude API, and the skill format is portable to other agents that read SKILL.md (Cursor, Codex, Gemini CLI). The forensic-read script needs only Python 3 — no dependencies.
The skill ecosystem is flooded with production skills and abstract mental-model dumps. clear-eye is deliberately small and applied: three concrete modes of skeptical thinking, each with a clear trigger, a structured output, and — where it helps — a script that does real work.
forensic-read— skepticism, with an evidence scannersecond-order— consequence-mappingpre-mortem— adversarial foresight
Looking for cross-cultural content adaptation? That's a different job — see the standalone vibe-translator skill.
MIT — use them, fork them, build on them.