6 projects
prompt-portability
Lint LLM chat-completion prompts for cross-provider portability issues
ragfresh
Detect stale, drifted, and ghost documents in RAG vector indexes
tooltrim
Drop-in compression for LLM agent tool outputs. Shrink bloated HTML/JSON/log results before they re-enter context — cut tokens, stay on-task, keep full output retrievable.
provensql
Sound-by-construction semantic diff for SQL: know whether a query edit changes the answer.
ctx-gate
LLM-agnostic context optimization proxy — reduce session token usage without losing accuracy
finagent-redteam
A red-team benchmark for AI agents in financial workflows: measures regulatory-control bypass (sanctions evasion, threshold structuring, dual-approval defeat) and the efficacy of guardrails that try to stop it.
