📄 Paper Reader + Heilmeier's Catechism

✨ What it does

Turns any research paper into a single Heilmeier-style analysis that is both a faithful summary and a critical review, with strict rules separating what the paper says from what the agent thinks.

🎯 What is different from other paper-reading skills

• 📝 Most tools just summarize. Useful for details, but it does not tell you whether the paper matters.
• 🔭 Significance comes first. What you usually need before digging in is a big-picture verdict.
• 🤖 Evaluation needs the agent's judgment, not just paraphrasing.
• 🚧 But judgment without guardrails kills trust. The line between paper and opinion must stay visible.
• 📦 This skill treats every paper as a product: structured evaluation framework, explicit per-question boundary, attribution markers on every opinion.

🧭 How it modifies the standard Heilmeier's Catechism

The Heilmeier's Catechism (after DARPA director George Heilmeier) is a checklist for evaluating proposed research projects. It asks: (1) what are you trying to do, no jargon; (2) how is it done today, what are the limits; (3) what is new and why will it succeed; (4) who cares; (5) what are the risks; (6) how much will it cost; (7) how long will it take; (8) what are the mid-term and final exams.

This skill repurposes it for completed papers:

• Q1 absorbs the one-sentence plain-language summary. No separate summary section needed.
• Old Q2 + Q3 → new Q2. Problem, state of the art, and limits are merged.
• Q3 absorbs the entire technical method, math included, and is strictly opinion-free.
• Old Q7 ("how long") is dropped. Papers report finished work.
• Old Q8 → "what are the experiments and results". The experiments are the exams; the results are the grades.

Two cross-cutting rules sit on top:

• 🎚️ Per-question opinion control. Q1 and Q3 are paper-only. Q4, Q5, Q6 invite analysis. Every personal judgment is prefixed with "In my opinion," or "My read is," so the source is never ambiguous.
• 🔗 Strict citation discipline. Every external reference must come from a web search performed in the same response. Citations from memory are forbidden. The only carve-out: repeating verbatim what the paper itself says about a work it cites.

⚠️ Known limitations

• 🆕 Very recent arXiv papers may not be retrievable until they are indexed.
• 🔒 Paywalled and subscription-only publishers cannot be downloaded. Works on open-access PDFs, arXiv, pasted text, and uploaded files.
• 🖼️ Scanned PDFs without an embedded text layer cannot be parsed. OCR is not part of the pipeline.