AI Security Corpus

5,660 links on AI security: prompt injection, LLM and AI agent security, MCP, coding agents, agent skills and supply chain, red teaming, guardrails, sandboxing, incidents and CVEs, governance.
_{One person's reading list that got out of hand. No gatekeeping: if it is even remotely about AI and/or security, it goes in. Added, not endorsed.}

Foundations · Attacks · Defense · Agents · Coding · Research · Practice · General · Stats · Maintenance

By the numbers

5,660

_links

29

_categories

108

_sections

570

_{arXiv papers}

1,155

_{GitHub repos}

570

_{posts on X}

Links per category

Largest sections

Section	Category	Links
Security News and Research	General Reading	748
AI Industry and Ecosystem	General Reading	613
Tools	Red Teaming	340
Resources	Supply Chain	315
Coding Tools and Assistants	Coding Tools	206

Recently added

2026-09-04 @office-agents/sdk - headless agent runtime SDK for AI-powered Office Add-ins
2026-09-04 Agent Development Kit - ADK - build multi-agent systems
2026-09-04 agent-browser - browser automation CLI for AI agents
2026-09-04 agents-cli - build AI agents on Google Cloud
2026-09-04 ANUS - Autonomous Networked Utility System open-source AI agent framework
2026-09-04 Azure Functions serverless agents runtime
2026-09-04 Centaur - self-hosted control plane for sandboxed, credential-safe AI agents in Slack
2026-09-04 CLI-Anything - making all software agent-native

mindmap
  root((AI Security Corpus · 5,660 links))
    Foundations · 297
      Frameworks and Standards · 95
      Governance and Policy · 138
      Threat Modeling · 48
      Architecture · 16
    Attacks · 612
      Prompt Injection · 117
      Jailbreaking · 26
      Model Attacks · 34
      Supply Chain · 315
      Incidents · 120
    Defense · 580
      Guardrails and Firewalls · 37
      Sandboxing and Isolation · 71
      Detection and Monitoring · 330
      Secrets Management · 95
      Honeypots and Deception · 29
      Anti-Crawling · 18
    Agents · 751
      Agent Security · 119
      Agent Identity · 126
      MCP - Model Context Protocol · 304
      Agent Frameworks · 202
    Coding · 455
      Secure Coding · 40
      Code Analysis · 186
      Coding Tools · 229
    Research · 515
      Papers · 402
      Benchmarks · 60
      Safety and Alignment · 53
    Practice · 624
      Red Teaming · 524
      Engineering Patterns · 48
      Privacy · 52
    General · 1,833
      General Reading · 1,833

🧱 Foundations

_{297 links across 4 categories}

Category	Links	Sections	What's inside
Frameworks and Standards	95	4	Risk frameworks, standards, and foundational security references
Governance and Policy	138	6	Compliance, AI policy, legal, and trust
Threat Modeling	48	4	Threat modeling frameworks, tools, and methodologies
Architecture	16	2	Secure AI design patterns and principles

💥 Attacks

_{612 links across 5 categories}

Category	Links	Sections	What's inside
Prompt Injection	117	7	Taxonomy, techniques, datasets, and defenses
Jailbreaking	26	1	Jailbreaking techniques and research
Model Attacks	34	5	Poisoning, backdoors, extraction, and adversarial ML
Supply Chain	315	1	Dependency attacks, model integrity, and signing
Incidents	120	5	Real-world breaches, exploits, and case studies

🛡️ Defense

_{580 links across 6 categories}

Category	Links	Sections	What's inside
Guardrails and Firewalls	37	3	Guardrails, firewalls, and runtime protection
Sandboxing and Isolation	71	3	Runtime containment and code execution security
Detection and Monitoring	330	4	Vulnerability scanners and threat detection
Secrets Management	95	3	Protecting secrets from AI agents
Honeypots and Deception	29	2	Honeypots and adversary engagement
Anti-Crawling	18	3	Tarpits, cloaking, data poisoning, and crawler access control

🤖 Agents

_{751 links across 4 categories}

Category	Links	Sections	What's inside
Agent Security	119	2	Agent-specific security concerns and threats
Agent Identity	126	1	OAuth, NHI, authentication, and authorization
MCP (Model Context Protocol)	304	12	Gateways, scanners, research, and tooling
Agent Frameworks	202	5	General agent frameworks, platforms, and tools

⌨️ Coding

_{455 links across 3 categories}

Category	Links	Sections	What's inside
Secure Coding	40	3	Rules files, vibe coding security, and secure prompt engineering
Code Analysis	186	3	SAST, code review, and vulnerability scanning
Coding Tools	229	2	IDE integrations, copilots, and assistants

📚 Research

_{515 links across 3 categories}

Category	Links	Sections	What's inside
Papers	402	8	Academic papers and surveys
Benchmarks	60	3	Evaluation frameworks and datasets
Safety and Alignment	53	2	AI safety, alignment, and privacy

🎯 Practice

_{624 links across 3 categories}

Category	Links	Sections	What's inside
Red Teaming	524	3	Offensive AI security, tools, and methodologies
Engineering Patterns	48	2	Harness engineering and building patterns
Privacy	52	4	Data leakage, PII protection, and exfiltration

📰 General

_{1,833 links across 1 category}

Category	Links	Sections	What's inside
General Reading	1,833	5	Blog posts, talks, opinion, commentary, and security news

Lists and Aggregators

System Prompts

System Prompts and Models of AI Tools

Prompt Design

How it is maintained

Every category is a plain markdown file of - [Title](url) lines under ## sections, so the whole thing greps, diffs, and forks cleanly.

What	How
Add links	`/add-resource <url> [url2] ...` — a Claude Code command that fetches titles, classifies each link into the right file and section, and commits
Search	`/search-resources <query>` — grep across every category file
Obsidian vault	`python3 scripts/sync-vault.py` regenerates `vault/` with one note per category, wiki-linked and tagged
Stats and charts	`python3 scripts/stats.py` recounts the corpus, redraws the SVGs in `assets/`, and rewrites the numbers above (also runs in CI on every push)

_{Links are added, not endorsed. Some of them are wrong, some are marketing, some are exploits. That is the corpus.}

_{Licensed CC BY 4.0: fork it, grep it, republish it, just link back.}

↑ back to top

ai-security-corpus