The Agent Skills Directory

[COMMAND_EXECUTION]: Executes local shell commands to navigate the project structure, manage local configuration in .claude-plugin/, and use Node.js for JSON parsing.
[EXTERNAL_DOWNLOADS]: Communicates with Adobe's official AEM Admin API (https://admin.hlx.page) to retrieve site configurations and project details. This is an expected behavior for an Adobe-authored management tool.
[DATA_EXFILTRATION]: Accesses the local IMS token file at ~/.aem/ims-token.json to authenticate API requests to Adobe services. While this involves sensitive credential access, it is performed by the vendor's own skill to interact with its official infrastructure.
[PROMPT_INJECTION]: Subject to Indirect Prompt Injection (Category 8). The skill fetches site metadata from an external API and interpolates it into the generated documentation without explicit sanitization.
Ingestion points: Site configuration data retrieved from https://admin.hlx.page/config/ (documented in Phase 1.1).
Boundary markers: Absent; API data is placed directly into markdown templates for the admin guide.
Capability inventory: The skill can write files (Write), execute shell commands (Bash), and invoke other skills (Skill).
Sanitization: No sanitization or validation of the external API content is performed before generating the ADMIN-GUIDE.md file.

handover-admin