The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill accesses a sensitive authentication token from the user's local filesystem at ~/.aem/ims-token.json. This token is required to authenticate requests against the AEM administration infrastructure for the purpose of validating project settings.
[EXTERNAL_DOWNLOADS]: The skill makes network requests to admin.hlx.page (Adobe's Helix/Edge Delivery Services infrastructure) to verify project organization identifiers.
[COMMAND_EXECUTION]: The skill relies on shell commands (bash, git, curl) and inline Node.js scripts to perform environment checks, parse configuration files, and manage directory structures within the project workspace.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data that could potentially be influenced by an attacker, specifically through Git repository URLs and user-supplied organization names or preview URLs.
Ingestion points: Ingestion occurs via the output of git remote get-url origin in Step 1.5.2 and through direct user input for the organization name or URL in Step 1.5.3.
Boundary markers: None are defined; the skill directly processes the input strings.
Capability inventory: The skill possesses Bash (file removal, creation, and network requests) and Write capabilities.
Sanitization: The skill uses basic string manipulation (cut, awk, grep) to parse identifiers from URLs, which provides limited protection against sophisticated command injection if the underlying shell environment is improperly handled.

handover