The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses persona instructions, casting the AI as an "elite Enterprise Software Architect." It also contains a behavioral override mandating the use of the "Superpowers plugin" and instructing the agent to prefer its methods over default behaviors.
[DATA_EXFILTRATION]: The discovery phase systematically scans and reads high-sensitivity project artifacts, including database schema dumps, migration history, and infrastructure configuration directories such as .github, deploy, and docker. While necessary for an audit, this provides a comprehensive view of internal architecture.
[PROMPT_INJECTION]: (Indirect Prompt Injection)
Ingestion points: The skill reads untrusted content from CLAUDE.md, AGENTS.md, README.md, and other documentation files within the audited project.
Boundary markers: No boundary markers or isolation instructions are provided to ignore commands embedded in audited files.
Capability inventory: The agent is granted broad read access to the entire project source code and write access to generate reports.
Sanitization: The skill does not implement any sanitization or validation of the content it audits.
[NO_CODE]: The skill consists entirely of instructional markdown and does not include any executable scripts or binary files.

implementation-status-auditor