Thomas Piacentini
← Workshop

Project

Skill Supply-Chain Scanner

Original text · English480 min

A skill-change review assistant quoting newly introduced capabilities.

Source evidence uses byte offsets. A line containing café occupies more UTF-8 bytes than visible letters, so character indices cannot safely slice a Rust string. Preserve start and end offsets while splitting lines.

Start with a learner workspace

Data management, Security and secrets audit. Language foundation: Rust ownership and Result.

Install Rust (rustc) and Python 3.10+. The core uses the Rust standard library; Python adapters compile into private temporary directories.

bash
python3 scripts/project_test.py skill-scanner --init learning-artifacts/skill-scanner
python3 scripts/project_test.py skill-scanner --stage 1 --path learning-artifacts/skill-scanner

Build route

  1. Retain byte-accurate source spans
  2. Detect named advisory patterns
  3. Score distinct evidence without inflation
  4. Build an explicit review gate

Run with your own inputs

After completing the stages, these commands run your workspace code on the original Orchard examples. Replace the sample paths with your own files.

bash
rustc --edition=2021 learning-artifacts/skill-scanner/cli.rs -o learning-artifacts/skill-scanner/scanner
learning-artifacts/skill-scanner/scanner projects/skill-scanner/examples/reviewable 3

To inspect the complete reference first, replace learning-artifacts/skill-scanner with projects/skill-scanner/solution in the same command. JSON results use schema_version: 1; paths and argument examples are explicit so another tool can consume them.

Integration boundary

The scanner emits versioned JSON with file paths, exact UTF-8 byte spans and advisory rules. Explicit patterns can miss obfuscation and flag benign documentation. below-threshold does not mean safe; scan status never replaces installation integrity or human review.

bash
python3 scripts/project_test.py skill-scanner --all --solution --strict
python3 scripts/project_test.py skill-scanner --all --path learning-artifacts/skill-scanner --strict

The first command checks the reference. The second checks your implementation. The published examples and tests are regression evidence, not a production certification or an unseen benchmark.

Primary references

Code and artifacts 33
What you take away

Saved in your notebook on this device.

Sources and limits

Reading checkpoints are self-reported. Test results are recorded separately and apply to the executed cases.

AI Engineering from Scratch · 3f9715c · MIT

Download this file ↓