kigland 5f2d087c0e feat: add LLM-generated content mode for poisoned skill files ...

When api_key is provided, uses GPT-4 to generate realistic skill
definitions with proper structure, responsibilities, workflow, and
constraints. The trigger is embedded under a randomized section label
(Runtime Configuration / Environment Bootstrap / Session Initialization
/ Workspace Calibration) instead of the conspicuous <diagnostic> tag.

Added 2 new camouflage profiles (devops, database) to the existing 4.
Falls back to hardcoded templates when no api_key is given.

2026-03-10 01:11:31 +08:00

..

docs

Create SOUL_PROMPT.py

2026-03-08 17:11:24 +08:00

method1_naive.py

传统的语义注入，通过角色扮演和忽略前文尝试绕过 SOUL.md

2026-03-08 01:32:04 +00:00

method2_cma_es.py

feat: add tool-call fitness evaluation for direct RCE verification

2026-03-08 22:29:08 +08:00

method3_honeypot.py

feat: add LLM-generated content mode for honeypot pages

2026-03-10 01:10:41 +08:00

method4_skills.py

feat: add LLM-generated content mode for poisoned skill files

2026-03-10 01:11:31 +08:00