Title: [High] No timeout on target agent or LLM HTTP requests; one stuck call blocks entire AI red-team run
Labels: bug, reliability, high, attack-runner, llm
Body:
Problem
All fetch() calls to the target agent and LLM providers are made without AbortSignal or timeout. A slow or hanging target or LLM API can block indefinitely.
Why this matters
AI red-team runs are long and sequential in places. One hung request freezes progress — no partial report from that point until manual kill. This is a core real-time operational failure mode.
Impact
High — Full run blocked; wasted time and unclear state.
Files / location
lib/attack-runner.ts — target fetch, loginForToken()lib/llm-provider.ts — OpenAI, Anthropic, OpenRouter
Title: [High] No timeout on target agent or LLM HTTP requests; one stuck call blocks entire AI red-team run
Labels:
bug,reliability,high,attack-runner,llmBody:
Problem
All
fetch()calls to the target agent and LLM providers are made withoutAbortSignalor timeout. A slow or hanging target or LLM API can block indefinitely.Why this matters
AI red-team runs are long and sequential in places. One hung request freezes progress — no partial report from that point until manual kill. This is a core real-time operational failure mode.
Impact
High — Full run blocked; wasted time and unclear state.
Files / location
lib/attack-runner.ts— targetfetch,loginForToken()lib/llm-provider.ts— OpenAI, Anthropic, OpenRouter