The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The script detect_wp_project.mjs reads wp-config.php, which typically contains highly sensitive database credentials and authentication salts. Although the implementation only extracts specific configuration flags using regular expressions, accessing the file in its entirety represents a data exposure within the skill's execution environment.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests file content and metadata from the scanned repository to generate a report with recommendations. This creates a surface where a malicious repository could attempt to influence the agent's behavior through the generated output.\n
Ingestion points: detect_wp_project.mjs recursively reads various files including block.json, package.json, and .php headers to detect project signals.\n
Boundary markers: The JSON report generated by the script does not include specific boundary markers or warnings to the agent regarding the untrusted nature of the scanned data.\n
Capability inventory: The agent uses the report's recommendations and commands fields to determine its next steps (e.g., running build or lint scripts), providing a path from untrusted file content to action.\n
Sanitization: The script does not perform specific sanitization or validation of the strings extracted from repository files before including them in the final report.

wp-project-triage