القائمة
Validate paper note links against actual paper references and influence
يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.
اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
يتجاوز الأمر المباشر Prompt المخصّص للمراجعة. افحص المصدر قبل تشغيله.
npx skills add https://github.com/D0miH/obsidian-second-brain-research-skills --skill validate-paper-linksيبقى الأمر في سطر واحد. مرّر أفقيًا لمراجعته كاملًا قبل النسخ.
تفضّل نسخة محلية؟ نزّل الملفات المتاحة حاليًا لدى SkillsMP.
عرض SKILL.md
استنادا إلى تصنيف SOC المهني
| name | validate_paper_links |
| description | Validate paper note links against actual paper references and influence |
Validate all [[Paper Title]] links in a paper note against the actual paper's references and citation patterns.
Goal: Prevent:
: not exchanged to -)A link should exist only if:
: as - (e.g., Attention - Is All You Need)A link should be removed if:
Read .claude/vault_maintanance.md and note the validate_paper_links: timestamp.
Default: incremental — only validate papers modified after that timestamp.
Convert timestamp (DD-MM-YYYY HH:MM) to YYYYMMDDHHMI format.
Run (from vault root): touch -t YYYYMMDDHHMI /tmp/validate_ref && find "Reading/" -newer /tmp/validate_ref -name "*.md"
If no papers were modified: report "All paper links already validated." and stop. Only run full rescan if user explicitly requests it or no timestamp exists yet.
This produces a list of paper file paths to validate.
For each paper file path from step 0, spawn one subagent using the Agent tool. Send all subagent calls in a single message (parallel execution).
If there are >5 papers: process in batches of 5 — send one batch, wait for results, then send the next.
Pass each subagent the PER-PAPER SUBAGENT PROMPT below with the paper's file path filled in.
Wait for all subagents to return. Merge their structured outputs into a single report grouped by severity across all papers:
🔴 CRITICAL (remove immediately):
🟡 NEEDS FIXING:
: not converted to -)🟢 OK:
❓ NEEDS USER JUDGMENT:
[[Paper Title]] lines from "🔗 Connections to Other Work": → - conversionAfter completing validation (regardless of whether fixes were applied):
date via Bash to get the current system time.claude/vault_maintanance.mdvalidate_paper_links: line to current date and time in format DD-MM-YYYY HH:MMpapers_since_validate_links: to 0For each paper, pass this prompt to the Agent tool (replace <PAPER_FILE_PATH> with the actual path):
You are a paper link validator for an Obsidian research vault.
Vault root: {{VAULT_ROOT}}
Paper to validate: <PAPER_FILE_PATH>
STEP 1: Read paper note
- Read the paper note at the given path
- Extract: URL from frontmatter (must be arxiv.org/html/ format), paper title, authors, year
- Extract all [[Paper Title]] links from the "🔗 Connections to Other Work" section
STEP 2: Fetch references via extraction script
Run (from vault root):
python3 .claude/skills/validate_paper_links/extract_refs.py <url-from-frontmatter>
- Accepts arxiv.org/html/ or arxiv.org/abs/ URLs (converts automatically)
- Returns a JSON array of reference titles — zero AI tokens consumed
- Strategy 1 (HTML/LaTeXML): fast, preferred
- Strategy 2 (source tarball fallback): for old papers where HTML unavailable
- If script fails entirely (exit code 1): fall back to WebFetch with prompt "Extract ONLY bibliography/references section titles."
STEP 3: Validate each [[link]] in the paper note
Check 1 — Reference exists?
- Does this paper appear in the bibliography JSON? (match case-insensitively)
- If NO → FLAG CRITICAL: "Not referenced in paper"
Check 2 — Influence level (only if passes Check 1)
- Citations ≥2 in this paper = influential ✓
- Cited once but in abstract/intro = influential ✓
- Cited once only in methods/results = LOW influence → FLAG CRITICAL
- Borderline → FLAG JUDGMENT
Check 3 — Title formatting
- Is it Title Case matching the reference?
- Are colons `:` in reference converted to `-` in link?
- If mismatch → FLAG FIX_NEEDED with corrected form
Check 4 — File name validity
- Link text should match exact file name (`:` → `-`)
- If mismatch → FLAG FIX_NEEDED
STEP 4: Return findings in EXACTLY this format (nothing else):
PAPER: [[filename-without-extension]]
CRITICAL: [[link]] | reason
FIX_NEEDED: [[link]] | reason | corrected form: [[correct link]]
OK: [[link]]
JUDGMENT: [[link]] | reason
Use one line per link. If a category has no entries, omit that category header entirely.
If all links are OK, return only: PAPER: [[filename]] / ALL_OK
extract_refs.py independently — zero AI tokens for reference extractionAvoid:
✅ Did I find modified papers before spawning subagents? ✅ Did I spawn one subagent per paper (in parallel where possible)? ✅ Did I collect and consolidate findings from all subagents before presenting? ✅ Did I group findings by severity across all papers? ✅ Did I get user approval before removing or fixing links? ✅ Did I update vault_maintanance.md with current timestamp?