The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill generates templates that include commands to inspect application configuration files which may contain database connection strings.\n
Evidence: references/database_ops_template.md includes a command to read appsettings.Development.json to check for connection settings.\n- [COMMAND_EXECUTION]: The skill creates executable bash scripts dynamically through variable substitution in predefined templates.\n
Evidence: SKILL.md workflow specifies loading templates from references/ and substituting {{VARIABLE}} placeholders to write new command files.\n- [PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection because it interpolates data from an external worker into executable scripts.\n
Ingestion points: Variables provided by ln-750 worker are processed in SKILL.md.\n
Boundary markers: No boundary markers or 'ignore' instructions are present in the templates.\n
Capability inventory: The skill writes to .claude/commands/ which contains shell script blocks executable by the agent.\n
Sanitization: There is no validation or escaping of the variable values before they are placed into bash script blocks.\n- [EXTERNAL_DOWNLOADS]: The documentation includes a visual diagram that loads a resource from a remote CDN.\n
Evidence: diagram.html fetches the Mermaid.js library from the well-known cdn.jsdelivr.net service for rendering workflow graphics.

ln-751-command-templates