Polito Notes

Convert PDF lecture slides into comprehensive bilingual (IT+EN) markdown notes for Polito university courses. Use when the user sends a PDF and specifies a c...

Security Scan

The skill's instructions generally match its stated purpose (turn PDFs into bilingual markdown notes), but there are notable mismatches and privacy risks: it references local tooling and specific filesystem paths that are not declared in metadata, and it explicitly requires preserving verbatim PDF content (which can include sensitive data).

Purpose & Capability

The skill's stated purpose (convert lecture PDFs into Italian and English markdown notes and place them in a course folder) aligns with the instructions to extract text, generate markdown, and write files under ~/Documenti/github/polito. However, the SKILL.md also references specific local components (e.g., ~/.local/share/local-rag/venv/ and ~/.openclaw/workspace/skills/lookupmark-local-rag/...) that go beyond a simple PDF-to-markdown converter and are not declared in the skill metadata (no required config paths or env vars). This is an unexplained dependence on local tooling and paths.

Instruction Scope

Instructions instruct the agent to read the user's filesystem (check/resolve ~/Documenti/github/polito folders), run pdftotext or a python fallback, create/overwrite files, and create backups. They also require 'Zero information loss' and explicitly state that PDF content will be preserved verbatim, which increases the risk of capturing sensitive or personal data. The instructions also reference running a local query script under ~/.openclaw/workspace/... for immediate search; these filesystem/behavioral expectations are broader than what the skill metadata declares and grant the agent broad discretion to read and write user files.

Install Mechanism

This is an instruction-only skill with no install spec and no bundled code — lowest install risk. The SKILL.md lists runtime dependencies (pdftotext, python3, and an optional local-rag venv), but there is no automated installer or remote download. That reduces supply-chain risk, though it does assume the user has particular tools and local paths available.

Credentials

The skill declares no required environment variables or config paths in metadata, yet the runtime instructions rely on specific local paths (~/.local/share/local-rag/venv, ~/.openclaw/workspace/skills/...) and the user's personal notes repository (~/Documenti/github/polito). This mismatch means the skill expects access to private filesystem locations without declaring them. Also, the 'Zero information loss' rule means potentially sensitive content from PDFs will be preserved and written to disk; that is a high privacy burden relative to the simple conversion task.

Persistence & Privilege

The skill is not set to always:true and does not request perpetual presence. It instructs the agent to write files into the user's notes repo and to create backups (notes.md.bak), which is consistent with its purpose. It does reference automatic pickup by a separate local-rag skill during indexing, but that is a usage note rather than an escalation of privileges within this skill.

What to consider before installing

This skill appears to do what it says (convert PDFs into bilingual markdown files), but it expects access to specific local folders and tools that the metadata doesn't declare. Before installing/using it: 1) Confirm you want the agent to read and write under ~/Documenti/github/polito (it will list folders, infer lecture numbers, create/overwrite notes and .bak files). 2) Be aware of the 'Zero information loss' rule — sensitive or personal data inside PDFs will be copied verbatim into the notes. If that is a concern, do not process sensitive documents or ask the author to redact them first. 3) The skill references local-rag and an ~/.openclaw workspace path; verify whether those paths exist and whether you want automatic indexing/search of generated notes. 4) Because this is instruction-only (no bundled code), make sure pdftotext and python3 are installed locally; test first with a non-sensitive PDF to confirm behavior. 5) If you want stronger safety: request the skill author declare required config paths and offer an option to redact personal data or require explicit user confirmation before writing/overwriting files.

Polito Notes Pipeline

Convert a PDF into two markdown files: Italian (notes.md) and English (notes-en.md), placed in the correct course folder.

Repository Structure

~/Documenti/github/polito/
├── first-year/
│   ├── first-semester/<course>/notes/
│   └── second-semester/<course>/notes/
├── second-year/
│   └── first-semester/<course>/notes/

Existing courses (check with ls — new ones may exist):

• 1st year 1st sem: architetture-dei-sistemi-di-elaborazione, big-data-processing-and-analytics, computer-network-technologies-and-services, data-science-and-database-technology
• 1st year 2nd sem: machine-learning-and-pattern-recognition, programmazione-di-sistema, software-engineering, web-applications-1
• 2nd year 1st sem: advanced-machine-learning, deep-natural-language-processing, large-language-models, robot-learning

Input

The user provides:

1. A PDF file (via file send or local path)
2. The course name (in Italian or English)
3. Optionally: lecture number and/or date

Output

Two files in <course>/notes/<N-kebab-title>/:

• notes.md — Italian
• notes-en.md — English translation (not a literal translation: adapt phrasing to natural academic English while preserving identical structure and technical accuracy)

Folder Naming

N-titolo-in-kebab-case — N is the lecture number (provided by user or inferred from content/prior folders).

Template

Both files follow this exact structure:

# [Title]

> **Course:** Course Name  
> **Lecture:** N  
> **Date:** YYYY-MM-DD  
> **Source:** original-filename.pdf

## Overview

[2-3 sentences: what this lecture covers and how it connects to previous topics]

## Content

### [Section 1 — Name]

[Fluid narrative prose. NOT bullet-list dumps. Explain each concept with
connecting logic, context, and motivation before diving into details.]

[Formulas in LaTeX inline/block, each symbol explained verbally right after.]

[When a comparison, taxonomy, or parameter set appears → **table**]

[When a process, pipeline, architecture, or relationship between concepts
appears → **Mermaid diagram**]

### [Section 2 — Name]
...

## Key Concepts

| Concept | Definition | Formula / Note |
|---------|-----------|----------------|
| ...     | ...       | ...            |

[Only at the end, as quick-reference. Does NOT replace the full explanations.]

## Connections

[Links to other lectures or courses when relevant.]

Rules

1. Zero information loss — every definition, formula, example, use case, caveat, and technical detail from the PDF must appear in the output. Nothing gets skipped.
2. Narrative flow — write in continuous prose, not mechanical lists. Bullet points only when genuinely natural (e.g., listing properties). Prefer "The key idea behind X is..." over "• X is...".
3. Visual reconstruction — every diagram, schema, or figure in the PDF becomes a Mermaid diagram when possible (flowcharts, sequences, hierarchies). For complex visual layouts (matrices, heatmaps, scatter plots, mathematical plots), use a descriptive paragraph explaining what the figure shows instead of forcing an inaccurate Mermaid diagram.
4. Formulas — LaTeX $$block$$ or $inline$. After each formula, explain every symbol in words.
5. Terminology — first occurrence of a technical term: bold. Standard English terms stay in English even in the Italian version (with Italian explanation on first use).
6. Update vs create — if the PDF covers topics already in an existing notes file, integrate the new content into the existing file. Don't duplicate.
7. Both files must be structurally identical — same sections, same tables, same diagrams. Only language differs.
8. Backup — before overwriting an existing notes file, copy it to <filename>.bak (e.g., notes.md.bak, notes-en.md.bak).
9. Sensitive data warning — PDFs may contain sensitive or personal data. All content from the PDF will be preserved verbatim in the generated markdown notes. Ensure notes are stored securely and not shared publicly.

Workflow

1. Extract text from PDF (use pdftotext or python script)
2. Identify the course and resolve the target folder
3. Check existing folders to determine lecture number
4. Generate notes.md (Italian) following the template
5. Generate notes-en.md (English) — same structure, natural academic English
6. Write both files to the target folder
7. Confirm to the user with: course, lecture number, folder path, and a brief summary of what was covered

RAG Integration

After generating notes, the new markdown files will be automatically picked up by the local-rag skill during the next indexing run (daily cron at 9 AM). No manual action needed.

To search across all notes immediately:

~/.local/share/local-rag/venv/bin/python ~/.openclaw/workspace/skills/lookupmark-local-rag/scripts/query.py "attention mechanism in transformers" --top-n 5

Example Output Structure

polito/second-year/first-semester/large-language-models/notes/
├── 01-introduction-to-llms/
│   ├── notes.md          # Italian
│   └── notes-en.md       # English
├── 02-transformer-architecture/
│   ├── notes.md
│   └── notes-en.md
└── ...