The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/scan-lwc-components.sh script is vulnerable to shell command injection. The script processes user-supplied query strings using an echo command inside a subshell expansion: $(echo "$QUERY" | tr ... ). If the query string contains shell metacharacters like $() or backticks, the bash shell will evaluate and execute them as commands during the variable expansion phase.
[COMMAND_EXECUTION]: The scripts/resolve-component-instructions.sh script contains a similar command injection vulnerability. It evaluates the component definition argument using echo inside a subshell: $(echo "$COMPONENT" | sed ... ). Maliciously crafted component names can trigger arbitrary command execution on the host system.
[EXTERNAL_DOWNLOADS]: The skill installs the official Salesforce CLI templates plugin to support FlexiPage generation tasks. This is a reference to a well-known service functionality.
[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves transforming natural language user intent into executable Salesforce metadata and Lightning Web Component (LWC) source code, which represents an attack surface for indirect prompt injection.
Ingestion points: User utterances describing requested components and page configurations are parsed to identify intents in Step 1 of the Adding Components workflow.
Boundary markers: There are no explicit instructions for the agent to use delimiters or ignore embedded instructions when interpolating user intent into tool arguments or generated files.
Capability inventory: The skill uses the Bash tool to execute sf project deploy (metadata deployment), sf template generate (file generation), and performs local file system write operations for LWC generation.
Sanitization: The skill does not provide mechanisms for validating or escaping user-provided names, labels, or descriptions before they are used in CLI commands or generated source code, potentially allowing a user to inject malicious metadata or code properties.

platform-flexipage-generate