The Agent Skills Directory

[PROMPT_INJECTION]: The skill provides templates and guidance for implementing defensive prompt engineering in AI agents. Specifically, it includes instructions to ignore user attempts to override system rules and prevents the disclosure of internal configurations, system prompts, or conversation summaries. These patterns are documented as hardening techniques to protect the integrity of the agents being developed.
[COMMAND_EXECUTION]: The skill instructs the agent to use standard Salesforce CLI (sf) commands for managing metadata, validating code, and deploying configurations. These commands are correctly used within the context of the Salesforce development lifecycle and leverage the --json flag to ensure safe, programmatic processing of results.
[SAFE]: Analysis of the skill's 74 files confirms that all behaviors and instructions are consistent with its stated purpose. The skill promotes secure practices such as using Named Credentials for API calls, enforcing USER_MODE in database queries, and adhering to the principle of least privilege through Permission Set management.

developing-agentforce