The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run a curl command to a local notification endpoint. Evidence: curl -s -X POST http://localhost:31337/notify in SKILL.md. This facilitates local system feedback for the user during query execution.
[COMMAND_EXECUTION]: Standard setup instructions are provided for adding the Signal MCP server via various AI tool CLIs. Evidence: Commands for hermes, claude, and codex in SKILL.md pointing to https://mcp.returnonsecurity.com/mcp. These refer to the legitimate service provider.
[DATA_EXPOSURE]: The skill attempts to load user-specific preferences from a designated local directory. Evidence: ~/.claude/LIFEOS/USER/CUSTOMIZATIONS/SKILLS/SecurityMarketData/ in SKILL.md. This is a standard practice for personalizing agent skill behavior.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external cybersecurity market intelligence, which represents an attack surface for indirect prompt injection. Evidence Chain: (1) Ingestion points: Data retrieved via signal-mcp tool calls and the REST fallback API. (2) Boundary markers: Not explicitly defined in the instructions. (3) Capability inventory: Subprocess calls via curl for notifications. (4) Sanitization: No explicit validation logic is defined, though the instructions mandate clear labeling of observed records versus analytical scores. This risk is inherent to the skill's primary purpose and is managed by clear data provenance instructions.

SecurityMarketData