The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill generates "paste-ready" prompts for downstream agents (e.g., Builder, Beacon). These prompts incorporate information from the analysis scope. While this creates a path for instructions to influence subsequent agent actions, it is the primary intended function of the skill to facilitate multi-agent workflows for risk mitigation. Ingestion points: SCOPE phase input defines the analysis boundary and target system. Boundary markers: Generated prompts are enclosed in markdown code blocks to separate them from report content. Capability inventory: The skill can spawn subagents using codex and agy tools to perform parallel analysis. Sanitization: Input data is processed through the FMEA methodology, though explicit sanitization steps for prompt interpolation are not detailed.
[COMMAND_EXECUTION]: The multi recipe involves invoking external CLI tools (codex, agy) to run parallel failure-mode enumeration across different AI models. This command execution is a core architectural feature for cross-engine reliability analysis and is documented as the standard operating procedure for this recipe.

omen