The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input (such as organizational details and policy preferences) to generate structured HR documents, which presents a surface for indirect prompt injection attacks. 1. Ingestion points: User prompts for policy creation, tool recommendations, and productivity metrics as defined in SKILL.md and the workflow in references/remote-work-ops.md. 2. Boundary markers: There are no explicit instructions or delimiters in the skill files to isolate user input from the agent's core instructions or to prevent the agent from following commands embedded in the input data. 3. Capability inventory: The skill is designed to write files to the _workspace/ directory and utilizes the sequential-thinking MCP tool for multi-step reasoning processes. 4. Sanitization: The instructions do not define any validation, filtering, or escaping mechanisms for the user-provided data before it is incorporated into the final output files.
[SAFE]: The skill provides legitimate guidance on Korean labor laws and references official government resources for compliance, such as the Ministry of Employment and Labor (moel.go.kr) and the National Law Information Center (law.go.kr).

hr-operations