The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/fdk-errors-warnings.js executes the fdk validate command. This execution is performed using child_process.spawn with shell: false, which prevents shell injection vulnerabilities. This is a standard operation for the intended purpose of the tool.
[DATA_EXPOSURE]: The skill scans application source code for sensitive information such as hardcoded credentials and OAuth secrets (e.g., in scripts/oauth-config-usage.js). This behavior is a core security auditing feature of the skill. The skill does not perform any network operations to exfiltrate this data.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted application code during the audit process. This creates a potential surface for indirect prompt injection if malicious files are included in the audited project. However, the risk is mitigated by the use of deterministic Node.js scripts for most checks, which process the content as raw text rather than executable instructions for the LLM.

fw-review