The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to take 'any goal' (untrusted data) from a user and process it through multiple phases (discovery, generation, execution). If a malicious goal contains instructions aimed at the agent rather than the task, the agent might execute unintended actions.- Ingestion points: Processes arbitrary text from the user described as 'any goal' in SKILL.md.- Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the user-provided goals are present in this orchestrator.- Capability inventory: The skill has access to powerful tools including Bash, Skill (to invoke other tasks), and several MCP plugins (mcp__atlas-engine, mcp__plugin_prd_go).- Sanitization: The orchestration logic does not describe any sanitization or validation of the input goal before passing it to the discovery or execution phases.

go