The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the command npx skiller@latest apply to synchronize and apply changes made to agent rule files.
[EXTERNAL_DOWNLOADS]: Fetches and runs the skiller package from the NPM registry using npx to update rule mirrors.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by automatically persisting conversation-derived content into permanent agent rules without sanitization.
Ingestion points: Conversation history and session outcomes processed in SKILL.md.
Boundary markers: Absent; lessons are written directly to .agents/rules/*.mdc without delimiters or safety warnings.
Capability inventory: Includes file-write operations and command execution (npx skiller@latest apply) across the agent environment.
Sanitization: Absent; the skill lacks a mechanism to validate or escape content extracted from the conversation before it is saved as a rule.

game-memory