StealthyIPIAttack

Generate a meeting-ready progress snapshot MD for the current AgentDojo attack research. Scans live runs/, running benchmark processes, and the latest docs/*status*.md to produce a timestamped report following the nx_sweep_status.md template. Use when the user asks for a "progress report", "미팅 자료", "진행 상황 요약", or says something like "현재 어디까지 왔는지 정리해줘". **Requires a researcher argument** (ys/yj/wj) — output is written to the researcher's sub-directory.

Maintain a Karpathy-style LLM-wiki (https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f) over the accumulating progress_*.md files in /home/dgu_wj92/ysl/StealthyIPIAttack/progress. Supports three operations — ingest (fold a new progress file into entity/concept pages), query (answer a natural-language research question by reading relevant wiki pages and citing sources), lint (detect stale, contradictory, or orphan pages). Use when the user says "wiki ingest", "wiki query", "wiki lint", "지금까지 MELON 뚫은 공격 정리해줘", "wiki에 넣어줘", or similar.

Sync code changes from the live AgentDojo workspace ($LIVE) to the IPIAttack git mirror and commit them. Use when the user asks to "최신화", "git에 올려줘", "sync mirror", "mirror 최신화", or says the mirror is out of date. Read-only for runs/, logs/, reports/, figures/ (data artifacts never go to git).

Compute ASR/DSR/utility metrics for an AgentDojo run directory using the notebook case taxonomy. Supports per-suite breakdown and baseline diff. Use when the user asks "how did run X do" or "compare X vs Y".

Classify attack-failure traces (security=False) into ignored / refused / wrong_tool / utility_drop with representative quotes. Use when the user asks "why did attack X fail" or drills into a low-ASR run.

Generate a Marp slide deck (MD + HTML) summarizing an AgentDojo run — headline metrics, case taxonomy, failure buckets, cache health, and optional baseline diff. Use when the user asks for a slide/report of an experiment result.

Launch an AgentDojo benchmark run (model × attack × defense × suites) with automatic watcher, smoke/full modes, and DAMI v1 safety guards. Use when the user wants to kick off or rerun an experiment.