Pinned Loading

1. prompt-injection-benchmark prompt-injection-benchmark Public

   A reproducible prompt-injection benchmark that measures which defenses actually work: each payload is replayed against every defense, the report separates whether the agent COMPLIED from whether th…

   Python

2. agent-sandbox-escape agent-sandbox-escape Public

   Which agent-sandbox escapes the isolation flags people actually deploy really stop -- every escape executed for real, not modeled, on the host.

   Python

3. temporal-multi-agent temporal-multi-agent Public

   Production-shaped multi-agent platform on Temporal: a main orchestrator plans a task, fans out to specialized agents that run concurrently and self-critique, then reviews and synthesizes, all as du…

   Python

4. least-privilege-agent least-privilege-agent Public

   A clear policy for what an AI agent may do unsupervised, and what bounds the blast radius when it is wrong. Deny-by-default policy, purpose binding, provenance tagging, opaque handles and an egress…

   Python

5. roster-ingest-guarantees roster-ingest-guarantees Public

   Where a Kafka to Iceberg pipeline stops being exactly-once: Flink 1.20, Iceberg 1.10, Kafka 4.3 and Trino 478 in Docker Compose. Turning off exactly-once checkpointing changed nothing; an old savep…

   Python

6. connector-sync-guarantees connector-sync-guarantees Public

   Incremental sync on a modified-since timestamp is lossy by construction, measured against a world where every change is known. What a rate limit really costs, which losses no overlap can recover, w…

   Python