The Agent Skills Directory

[SAFE]: The skill instructions emphasize security best practices, specifically the redaction of sensitive information such as OIDC secrets, tokens, and credentials in any output generated by the agent. This is explicitly stated in the 'Privacy and Security' section of SKILL.md and echoed in examples/04-provider-operations-refusal.md.
[SAFE]: The skill uses npx skills add commands in its 'When NOT to Use' section to redirect users to appropriate specialized skills. These references (full-stack-skills/ddd4j-skills) are consistent with the skill's author context and represent standard modular skill management within the vendor's ecosystem.
[INDIRECT_PROMPT_INJECTION]: As a code analysis and implementation skill, it naturally possesses an attack surface where it ingests untrusted source code (POMs, Java source, tests). However, the skill provides instructions for the agent's behavior and does not include capabilities for arbitrary code execution, file writing, or network exfiltration that could be easily weaponized by injected content.
Ingestion points: Analyzes local source code and configuration files in the user's workspace as defined in references/source-evidence.md.
Boundary markers: The skill lacks explicit boundary markers for separating its instructions from the analyzed code, which is a common but low-risk surface for instructions aimed at developer guidance.
Capability inventory: Primarily involves generating code snippets, architectural advice, and test plans. No dangerous subprocess or network capabilities are defined.
Sanitization: Includes explicit rules to redact secrets and credentials from all output.

ddd4j-quarkus-auth