Defenses against indirect prompt injection in LLM agents. Focused on the retrieval surface: untrusted content from search, RAG, and tool outputs.
Deconvolute Labs
Production infrastructure for AI agent systems. Reliability, observability, cost efficiency.
Pinned Loading
Repositories
Showing 7 of 7 repositories
- interbolt Public
Provenance-gated tool calls for AI agents. A local YAML+CEL policy gates each tool call on the provenance of its arguments. Deterministic, in-process, no model.
- deconvolute Public
Policy-as-code enforcement and observability for MCP tool calls. Wraps AI agent sessions with cryptographic integrity checks, argument-level CEL policies, and a full audit trail.
- .github Public
- yaramint Public
Generate YARA rules automatically from positive and negative examples. For PII detection, secret scanning, and prompt injection.
- mcp-deconvolute-demo Public
Live PoC: MCP attacks that compromise AI agents mid-session and how to block them in a few lines of code.
- benchmarks Public
Reproducible security benchmarking for the Deconvolute SDK and AI system integrity against adversarial attacks.
- trojan-rag-demo Public
A demonstration of RAG poisoning attacks using dormant documentation injections.
Top languages
Loading…
Most used topics
Loading…