The Agent Skills Directory

[SAFE]: The skill implements a professional support workflow specifically for Contentstack products, using official documentation and references.
[SAFE]: The 'Security' section in the main instructions explicitly forbids the agent from requesting or exposing sensitive credentials like API keys, management tokens, or auth headers. It mandates the use of placeholders and redaction for logs.
[SAFE]: The toolset is restricted to read-only file access (Read, Grep, Glob), preventing destructive actions or unauthorized modifications to the user's environment.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it ingests and processes untrusted repository code (e.g., package.json, source files) via the Read tool.
Ingestion points: Repository files are read in Step 4 of the workflow to detect frameworks and rendering modes.
Boundary markers: The instructions lack explicit delimiter markers for processed code, though they provide high-level safety constraints for handling data.
Capability inventory: The skill is limited to read-only file inspection; it has no capabilities for writing files, executing shell commands, or making network requests.
Sanitization: The skill instructs the agent to redact secrets from output and use placeholders for environment variables, reducing the risk of accidental data leakage during processing.

cms-live-preview-visual-editor-support-assistant