메뉴
Parse Terminus2 agent trajectories. Terminus2 stores each episode as agent-logs/episode-N/response.txt (JSON with commands/analysis/plan) and the next episode's prompt.txt contains the observation (New Terminal Output section).
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
SOC 직업 분류 기준
Parse Claude Code trajectories from the terminal-bench recording format. Extracts tool-use steps from sessions/claude_code.log (LiteLLM proxy log containing full Claude API request JSON) and metadata from results.json. Captures thinking blocks, parallel tool calls, and the full Claude Code tool taxonomy.
Parse MiniSWE agent trajectories. MiniSWE stores data in one of three formats: agent-logs/mini.traj.json (JSON with messages array), sessions/agent.log (structured text log), or commands.txt (keystroke lists).
Parse OpenHands agent trajectories. OpenHands stores events as either sharded JSON files under sessions/sessions/*/events/ (or event_cache/) or as flat session JSON files under sessions/.
| name | terminus2 |
| description | Parse Terminus2 agent trajectories. Terminus2 stores each episode as agent-logs/episode-N/response.txt (JSON with commands/analysis/plan) and the next episode's prompt.txt contains the observation (New Terminal Output section). |
| fingerprints | ["agent-logs/episode-*/response.txt"] |
| priority | 50 |
| metadata | {"version":"1.0","source":"seed"} |
run_dir/
results.json # Task metadata
agent-logs/
episode-0/response.txt # JSON: {commands, analysis, plan}
episode-1/prompt.txt # Contains "New Terminal Output:" section
episode-1/response.txt
...
Presence of agent-logs/episode-*/response.txt files.