The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits a vulnerability surface for indirect prompt injection due to its requirement to process untrusted project data while having access to build and deployment capabilities. Ingestion points: The skill reads project source code, POM files, Git metadata, and build artifacts such as target/resume.properties, as specified in references/source-evidence.md. Boundary markers: The instructions lack explicit delimiters for untrusted data but enforce a 'read-only' posture for initial validation and require explicit user authorization for deployment actions. Capability inventory: The skill can execute Maven build commands and publish artifacts to private repositories. Sanitization: Instructions strictly mandate the redaction of tokens, cookies, OIDC secrets, and repository credentials from all output to prevent accidental exposure.
[COMMAND_EXECUTION]: The skill utilizes shell-based commands to perform build validation and artifact publication. Evidence: SKILL.md and workflow.md describe executing clean builds per maintenance line and performing serial deployments to private repositories. Safety: High-impact operations like deployment are restricted to authorized sessions and require verification of CI terminal success for the specific Git SHA.
[EXTERNAL_DOWNLOADS]: The documentation references external resources for extended functionality, specifically sibling skills from the same author. Evidence: SKILL.md suggests installing 'full-stack-skills/ddd4j-skills' using 'npx' for production hardening and testing purposes. Context: These are identified as vendor-owned resources belonging to the 'full-stack-skills' organization and do not represent unverified third-party risks.

ddd4j-javalin-release