The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions specifically target the bypass of external integrity systems like 'Antiplagiat' and 'Textovod'. It directs the agent to utilize anti-detection strategies, such as varying sentence lengths and introducing intentional linguistic 'imperfections' or 'authorial observations', to circumvent automated AI-content identification.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to instructions embedded in user-supplied thesis topics. Ingestion points: User prompts for thesis segments (SKILL.md). Boundary markers: Absent; the agent is not warned to ignore instructions within the user input. Capability inventory: The skill possesses no dangerous capabilities such as file system writes, network access, or command execution. Sanitization: No sanitization or input filtering is performed.
[SAFE]: No malicious scripts, exfiltration patterns, or persistence mechanisms were found. All referenced materials are stylistic markdown files.

master-thesis