The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves reading and analyzing external project files, which constitutes a potential injection vector if those files contain malicious instructions.
Ingestion points: The agent is instructed to read Maven pom.xml files, source code, and integration test results in SKILL.md and references/workflow.md.
Boundary markers: There are no explicit instructions for the agent to ignore or delimit instructions found within the analyzed project files.
Capability inventory: The skill allows for deep filesystem access within the workspace and can generate configuration changes based on analyzed data.
Sanitization: No specific sanitization or escaping mechanisms for external content are defined in the instructions.
[METADATA_POISONING]: The reference documentation contains a hardcoded absolute local path which exposes author-specific metadata.
Evidence: The file references/source-evidence.md specifies a root directory of /Users/wandl/workspaces/workspace-ddd4j/workspace-ddd4j-boot/ddd4j-boot. This discloses the author's local username and directory structure.
[COMMAND_EXECUTION]: The skill documentation suggests using npx to install related tools, which involves downloading and executing code.
Evidence: SKILL.md contains suggestions to run npx skills add full-stack-skills/ddd4j-skills --skill [name].
Context: These resources originate from the skill's own vendor, 'full-stack-skills', and are part of the standard installation workflow.

ddd4j-boot-observability