The Agent Skills Directory

[SAFE]: The skill incorporates a robust 'Quality Gate' phase that specifically instructs the agent to check for hardcoded secrets, API keys, and insecure CORS configurations in software projects before completion.
[COMMAND_EXECUTION]: The skill instructs the agent to perform standard development operations such as dependency installation (npm ci, pip install) and syntax validation as part of its quality assurance workflow.
[PROMPT_INJECTION]: The skill processes project-specific data from files like LESSONS.md and source code to influence its orchestration behavior. This ingestion of local file content represents a surface for indirect prompt injection, though it is mitigated by a structured workflow and requirements for user confirmation at each phase.
[SAFE]: The skill includes a 'Self-Improvement Protocol' that allows it to suggest updates via a SKILL-PATCH.md file, but it explicitly forbids autonomous modification of its own instructions and requires human review for all changes.

duru-conductor