The Agent Skills Directory

PROMPT_INJECTION (LOW): The skill facilitates the creation of rule files based on user input, which presents a surface for indirect prompt injection. Evidence: 1. Ingestion points: User requirements for new rules are processed during the creation flow (SKILL.md). 2. Boundary markers: The markdown templates provided in SKILL.md and REFERENCE.md do not use delimiters or warnings to isolate user-provided rule content. 3. Capability inventory: The skill is designed to perform file system write operations to save rules in the .claude/rules/ directory. 4. Sanitization: No sanitization or validation of the rule content is performed before it is written to the filesystem. This could allow a malicious user to store instructions that compromise the agent's behavior in future sessions.

rule-creator