llm-agent-security-principleslisted

Structural security design principles for building LLM agents, autonomous systems, and self-improving harnesses. Use when designing a new agent, reviewing agent architecture, adding a new capability, or auditing an existing agent for prompt injection resistance. Covers three load-bearing principles — Security by Absence, Untrusted Content Boundary, One External Adapter — grounded in the prohibition-strength hierarchy (absence > scaffolding enforcement > untrusted boundary), plus concrete defense patterns for HTTP, credentials, and LLM hosts.
shimo4228/llm-agent-security-principles · ★ 1 · AI & Automation · score 78

Install: claude install-skill shimo4228/llm-agent-security-principles

# LLM Agent Security Principles Web-app security (OWASP, auth, input validation) does not cover the threats that matter most for autonomous LLM agents: prompt injection, capability escalation through memory, and blast-radius accumulation. This skill is the structural playbook for those threats. The core insight: **you cannot reliably restrict an LLM's behavior at runtime. You must shape the system so the dangerous behaviors are structurally impossible.** Three principles do the heavy lifting. They are not independent — they are three axes of the same idea: minimize attack surface by construction. This skill is the "how" counterpart to four ADRs in the [Agent Attribution Practice (AAP)](https://github.com/shimo4228/agent-attribution-practice) research line (concept DOI [10.5281/zenodo.19652013](https://doi.org/10.5281/zenodo.19652013)): - [AAP ADR-0001 Security by Absence](https://github.com/shimo4228/agent-attribution-practice/blob/main/docs/adr/0001-security-by-absence.md) - [AAP ADR-0002 Deterministic Prohibition at the Scaffolding Layer](https://github.com/shimo4228/agent-attribution-practice/blob/main/docs/adr/0002-deterministic-prohibition-at-scaffolding.md) - [AAP ADR-0003 Untrusted Content Boundary](https://github.com/shimo4228/agent-attribution-practice/blob/main/docs/adr/0003-untrusted-content-boundary.md) - [AAP ADR-0004 Single External Adapter per Agent Process](https://github.com/shimo4228/agent-attribution-practice/blob/main/docs/adr/0004-single-external-ada