The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple sources, which presents an attack surface for indirect prompt injection.
Ingestion points: The skill ingests user queries, project context files (project_*.md), and a business portfolio file (portfolio.local.md). Additionally, it is designed to load custom advisor dossiers from a user-controlled configuration directory (${MAKERSKILLS_CONFIG:-$HOME/.config/makerskills}/maker-council/advisors/).
Boundary markers: While the skill provides a strict output template and "Grounding rules" to guide the agent, it lacks explicit delimiters (like XML tags or clear separators) to prevent instructions inside custom dossiers or context files from being interpreted as commands by the agent.
Capability inventory: The agent has the capability to perform network-based research via the deep-research and watch-video tools and write session archives to the local file system.
Sanitization: There is no evidence of automated sanitization or validation for the content of loaded dossiers or portfolio files beyond the natural language instructions provided to the agent.

maker-council