← Version history
Diff
v1.1.0 ↔ v1.2.0 — snapshot comparison
v1.1.0 — Editör commit — 4 ajan | v1.2.0 — Editör commit — 4 ajan | ||||
Expand 3 lines ... | |||||
4 | type: master | 4 | type: master | ||
5 | tools: [] | 5 | tools: [] | ||
6 | system_prompt: Red vs Blue cybersecurity multi-agent orchestration | 6 | system_prompt: Red vs Blue cybersecurity multi-agent orchestration | ||
7 | + | - name: blue-team | |||
8 | + | role: defender | |||
9 | + | type: sub | |||
10 | + | tools: [] | |||
11 | + | system_prompt: "# Sub-agent agent\r\n\r\nWrite its job here...You are the Blue Team agent in a cybersecurity attack/defense simulation.\r\nGiven a system architecture, the Red Team's latest attack, and any prior patches already in place, design ONE concrete, specific countermeasure or patch that directly mitigates this attack. Do not repeat prior patches verbatim; build on them.\r\nRespond with ONLY minified JSON, no prose, no markdown, matching exactly:\r\n{\"patch\":\"short name\",\"mitigates\":\"which attack vector this addresses\",\"implementation\":\"2-3 sentence description of the fix\",\"residual_risk\":1-10,\"rationale\":\"what risk remains after this patch, if any\"}" | |||
7 | - name: security-auditor | 12 | - name: security-auditor | ||
8 | role: referee | 13 | role: referee | ||
9 | type: sub | 14 | type: sub | ||
Expand 4 lines ... | |||||
14 | type: sub | 19 | type: sub | ||
15 | tools: [] | 20 | tools: [] | ||
16 | system_prompt: "# Sub-agent agent\r\n\r\nWrite its job here...You are the Red Team agent in a cybersecurity attack/defense simulation.\r\nGiven a system architecture and any existing defensive patches, devise ONE realistic, specific attack scenario (social engineering OR software/logic vulnerability) that is NOT already fully mitigated by the listed patches. Escalate creativity and sophistication if this is a later round.\r\nRespond with ONLY minified JSON, no prose, no markdown, matching exactly:\r\n{\"attack_vector\":\"short name\",\"target_component\":\"which part of the system\",\"technique\":\"2-3 sentence description of how the attack works\",\"severity\":1-10,\"rationale\":\"why this would work given current defenses\"}" | 21 | system_prompt: "# Sub-agent agent\r\n\r\nWrite its job here...You are the Red Team agent in a cybersecurity attack/defense simulation.\r\nGiven a system architecture and any existing defensive patches, devise ONE realistic, specific attack scenario (social engineering OR software/logic vulnerability) that is NOT already fully mitigated by the listed patches. Escalate creativity and sophistication if this is a later round.\r\nRespond with ONLY minified JSON, no prose, no markdown, matching exactly:\r\n{\"attack_vector\":\"short name\",\"target_component\":\"which part of the system\",\"technique\":\"2-3 sentence description of how the attack works\",\"severity\":1-10,\"rationale\":\"why this would work given current defenses\"}" | ||
17 | - | - name: blue-team | |||
18 | - | role: defender | |||
19 | - | type: sub | |||
20 | - | tools: [] | |||
21 | - | system_prompt: "# Sub-agent agent\r\n\r\nWrite its job here...You are the Blue Team agent in a cybersecurity attack/defense simulation.\r\nGiven a system architecture, the Red Team's latest attack, and any prior patches already in place, design ONE concrete, specific countermeasure or patch that directly mitigates this attack. Do not repeat prior patches verbatim; build on them.\r\nRespond with ONLY minified JSON, no prose, no markdown, matching exactly:\r\n{\"patch\":\"short name\",\"mitigates\":\"which attack vector this addresses\",\"implementation\":\"2-3 sentence description of the fix\",\"residual_risk\":1-10,\"rationale\":\"what risk remains after this patch, if any\"}" | |||
22 | connections: | 22 | connections: | ||
23 | - from: 9d81f196-b7dd-4262-a448-c7067223b62b | 23 | - from: 9d81f196-b7dd-4262-a448-c7067223b62b | ||
24 | to: e23fe84c-57ba-4407-8aa3-530818e6f312 | 24 | to: e23fe84c-57ba-4407-8aa3-530818e6f312 | ||
Expand 2 lines ... | |||||