The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the sf CLI (Salesforce CLI) to perform deployment validations. These commands are well-structured and use the --json flag to ensure that the AI processes the output as data rather than executable instructions.
[DATA_EXFILTRATION]: The skill maintains deployment state by writing job IDs to the .sfdx/last-validation.json file. This is a standard practice for Salesforce development tools and does not involve sending sensitive information to external or untrusted servers.
[REMOTE_CODE_EXECUTION]: No remote code execution or suspicious external downloads were detected. The skill uses a local script reference (sf-deploy-gate) and the official Salesforce CLI tools.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes validation errors and test results from a Salesforce org.
Ingestion points: Results from sf project deploy validate (specifically component failures and test messages).
Boundary markers: Absent.
Capability inventory: Shell execution (Bash) and file system access (Write).
Sanitization: None. While the surface exists, the risk is negligible as the data consists of technical error messages from a controlled environment (Salesforce).

platform-deploy-validate