The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes curl shell commands to send status notifications to a local endpoint at http://localhost:31337/notify. It also uses echo to log workflow execution data, including a summary of user input, to the local file ~/.claude/LIFEOS/MEMORY/SKILLS/execution.jsonl.\n- [PROMPT_INJECTION]: The TestIdea workflow ingests arbitrary user-provided ideas, strategies, or investments and processes them through an analysis pipeline involving multiple high-capability agents.\n
Ingestion points: The workflow in Workflows/TestIdea.md takes user input and passes it to parallel analysis agents, as well as the RedTeam, FirstPrinciples, and Council skills.\n
Boundary markers: The skill does not define explicit boundary markers or delimiters to isolate user-supplied content from the agent's instructions, which could allow the input to influence the behavior of the sub-agents.\n
Capability inventory: The skill orchestrates agents with access to Research and WebSearch capabilities and performs local shell command execution.\n
Sanitization: No sanitization or validation of the user input is performed before it is used to generate summaries or used in prompts for secondary agents.

WorldThreatModel