The Agent Skills Directory

[PROMPT_INJECTION]: The skill ingests untrusted writing samples to generate a systemPrompt that is persisted to the workspace and used in future creative writing tasks. This creates a surface for indirect prompt injection, where an attacker could embed hidden instructions within a text sample that are then incorporated into the agent's long-term behavior profile. \n- Ingestion points: Processes writing samples provided via manual paste or project files as described in SKILL.md. \n- Boundary markers: No delimiters or safety instructions are defined to separate writing samples from agent commands. \n- Capability inventory: Employs file:read for data collection, file:write for profile persistence, and network requests to an internal analyzer service. \n- Sanitization: No sanitization or filtering logic is present to validate the content of analyzed text before it influences the generated system prompt.

style-clone