The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes standard build and test commands (e.g., npm build, cargo test, make build) based on the detected project environment. These are expected behaviors for a build integration skill.
[EXTERNAL_DOWNLOADS]: The skill performs security scans using standard tools like npm audit, cargo audit, or snyk test. These operations involve connecting to official package registries or vulnerability databases, which are trusted sources for security auditing.
[PROMPT_INJECTION]: The skill contains instructional prompts (e.g., 'WAIT for user approval', 'Never narrate your internal process') that guide the agent's behavior during the build process. These are standard operational instructions and do not attempt to bypass safety filters or override system-level constraints.
[DATA_EXFILTRATION]: While the skill interacts with the local filesystem to read configuration and write build reports, it does not show any signs of exfiltrating sensitive data to external servers. Network operations are limited to health checks (curl localhost) and standard security audit tools.
[SAFE]: The skill follows best practices by requiring user approval at each critical step (e.g., before proceeding to deployment or after build failures) and maintains a clear audit trail in the project manifest.

aidlc-build