The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends several well-known and vendor-maintained libraries and CLI tools for content extraction and transformation tasks.
References Sanity-maintained tools such as contentful-to-sanity and @portabletext/markdown.
Suggests ecosystem-standard packages including fast-xml-parser, gray-matter, and jsdom.
Mentions official CLI tools for target and source platforms like contentful-cli, wp-cli, and strapi.
[COMMAND_EXECUTION]: Instructions guide the agent to perform network-based extraction and tool initialization using standard utilities.
Guides the use of curl for interacting with source CMS APIs (AEM, WordPress, Strapi) using environment variables for authentication.
Recommends initializing extraction and migration utilities via npx (e.g., npx contentful-to-sanity@latest).
[PROMPT_INJECTION]: The skill inherently handles a significant attack surface for indirect prompt injection given its primary purpose of content migration.
Ingestion points: Untrusted data enters the agent context via source CMS APIs, WXR/XML exports, Markdown files, and static HTML crawling across all reference guides.
Boundary markers: The skill relies on a documented "Agent Workflow" in references/general.md that enforces an "analyze, plan, generate" pattern where the user reviews all generated code before execution.
Capability inventory: The skill uses network operations (curl), tool execution (npx), and script generation for file system access (references/aem.md, references/strapi.md).
Sanitization: Recommends the use of specialized libraries like jsdom and @portabletext/block-tools to safely parse and transform untrusted HTML and rich text into structured Portable Text.
[SAFE]: Several operational security guardrails are integrated into the instructions.
Credential Handling: In references/contentful.md, the skill explicitly warns the agent to ask for user confirmation before accessing local configuration files like ~/.contentfulrc.json for tokens.
Control of Execution: The core stance in references/general.md prohibits silent migrations, requiring a human checkpoint before any data is written to the destination.

sanity-migration