The Agent Skills Directory

[PROMPT_INJECTION]: The skill utilizes a multi-step agent pipeline where findings from initial stages (e.g., market research) are directly interpolated into the prompts for subsequent agents (e.g., Business Analyst) using variables like ${marketData}. This design creates a surface for indirect prompt injection, as data generated by one agent could potentially contain instructions that influence the behavior of the next. However, this is a standard architecture for complex SOPs and is considered low risk in this context.\n
Ingestion points: Memory retrieval from market-research/product-launch-2024/analysis as seen in Phase 1.\n
Boundary markers: The skill does not use specific delimiters or XML tags to wrap interpolated data to separate it from instructions.\n
Capability inventory: The skill can spawn multiple specialized agents (Task function) and initialize a development swarm via MCP tools.\n
Sanitization: No explicit content filtering or verification is performed on data moving between orchestration steps.\n- [COMMAND_EXECUTION]: The workflow orchestrates the launch process using platform-specific abstractions, including the Task function for agent delegation and the mcp__ruv-swarm__swarm_init tool for initializing parallel developer swarms. These operations are within the scope of the skill's stated purpose of coordinating a large-scale product launch.

sop-product-launch