The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes local YAML files (menus and schemas) which serves as an ingestion point for potentially untrusted data that could contain malicious instructions.
Ingestion points: The skill identifies and reads existing configuration files in */menus/* and */schemas/* directories (SKILL.md).
Boundary markers: No specific boundary markers or instructions to ignore embedded prompts within the ingested files are defined.
Capability inventory: The agent is granted Bash, Write, and Edit tools (SKILL.md), providing a significant capability set if the agent were to be manipulated by instructions found in the data files.
Sanitization: Content from the discovered files is processed without explicit sanitization or validation steps.
[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes dynamic context injection to automate the discovery of project-specific configuration files during the skill's initialization.
Evidence: The SKILL.md file contains !find commands used to list existing menu and schema files at load time.
Context: These shell operations are benign discovery tools used for project context gathering and do not involve user-controlled input, sensitive file access, or network exfiltration.

infrahub-managing-menus