The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it is designed to read and process message content from external LINE users, which could contain malicious instructions.
Ingestion points: External messages are ingested into the agent context via the agent-line message list command and SDK-based listeners as documented in SKILL.md, templates/chat-summary.sh, and references/common-patterns.md.
Boundary markers: No explicit delimiters (e.g., XML tags or backticks) or "ignore embedded instructions" warnings are implemented in the provided templates or instructions to isolate external message text from the agent's core instructions.
Capability inventory: The agent has the capability to execute shell commands via the agent-line tool (configured in allowed-tools), perform network operations through the binary, and manage a persistent local memory file at ~/.config/agent-messenger/MEMORY.md.
Sanitization: There is no evidence of sanitization, filtering, or validation of the retrieved message content before it is presented to the agent or processed by the provided scripts.

agent-line