AI & Software EngineeringKnowledge Atlas
Catalogues / NDSE2 chapter guides and sources

C:C06

Chapter 6: Securing Generated Code and Agent Runtimes

A guide to the published second edition, with literal citations separated from wider editorial reading suggestions.

Open this group in the atlasChapter catalogue entry

Chapter guide

Threat-models two distinct surfaces: the generated application and the agent runtime. Dependency provenance, prompt injection, tool permissions and external services become enforceable boundaries, with security evidence tied to the release decision.

Mettere in sicurezza il codice generato e i runtime degli agent

Sintesi in italiano

Analizza due superfici di attacco distinte: l'applicazione generata e il runtime dell'agent. Provenienza delle dipendenze, prompt injection, permessi degli strumenti e servizi esterni diventano vincoli applicabili, collegando le evidenze di sicurezza alla decisione di rilascio.

Learning goals

  • threat-model the generated product and the agent runtime separately
  • enforce tool permissions independently of instructions embedded in files
  • specify the evidence required before accepting a security-sensitive patch
Obiettivi in italiano

Obiettivi di apprendimento

  • costruire threat model distinti per il prodotto generato e per il runtime dell'agent
  • far rispettare i permessi dei tool indipendentemente dalle istruzioni incorporate nei file
  • specificare le evidenze necessarie prima di accettare una patch sensibile dal punto di vista della sicurezza
Chapter outline
  • 6.1 Two attack surfaces, two threat models
  • 6.2 Functional correctness is not security
  • 6.3 What vulnerability studies actually measure
  • 6.4 Security-directed generation and remediation
  • 6.5 Dependency provenance and hallucinated packages
  • 6.6 Prompt injection through repository and tools
  • 6.7 Harness configuration, skills and external services
  • 6.8 A secure agent execution profile
  • 6.9 Threats turned into enforceable boundaries
  • 6.10 Security evidence and release decision
  • 6.11 Worked engineering example: two threats, two defenses
  • 6.12 Practice exercise: threat-model both the product and the agent
Struttura del capitolo
  • 6.1 Due superfici di attacco, due threat model
  • 6.2 La correttezza funzionale non basta per la sicurezza
  • 6.3 Che cosa misurano davvero gli studi sulle vulnerabilità
  • 6.4 Generazione e remediation guidate dalla sicurezza
  • 6.5 Provenienza delle dipendenze e package inventati
  • 6.6 Prompt injection attraverso repository e tool
  • 6.7 Configurazione dell'harness, skill e servizi esterni
  • 6.8 Un profilo di esecuzione sicura per gli agent
  • 6.9 Dalle minacce ai confini applicabili
  • 6.10 Evidenze di sicurezza e decisione di release
  • 6.11 Esempio ingegneristico svolto: due minacce, due difese
  • 6.12 Esercizio pratico: costruire due threat model, per prodotto e agent

Cited in the published chapter

10 linked atlas references from literal citations in the English and Italian editions. General bibliography membership and identifier-index entries are counted separately.

SE@PoliMi bibliography 9

Martin Fowler’s website 1

Trace literal citations to chapter sections

Locations come from the supplied English and Italian source files. Chapter text and figure captions are identified separately; identifier-index references and general bibliography membership are not treated as chapter citations.

Complete citation locators and source hashes

Editorial reading mappings

Broader topical suggestions can overlap the cited sources. A mapping does not add a citation to the published edition or establish that a practice is effective.

SE@PoliMi bibliography 41

Addy Osmani 2

The Pragmatic Engineer 5

Martin Fowler’s website 1

Earlier Non-Deterministic books 14

Software Engineering Practices 10