The Agent Skills Directory

[SAFE]: The skill serves as a development aid for Salesforce Agentforce Agent Script DSL. Detailed analysis of all 81 files found no evidence of malicious intent or security vulnerabilities.
[PROMPT_INJECTION]: Deterministic detector hits in references/fsm-architecture.md are false positives. The file contains a table describing Salesforce's standard platform topics and includes example injection phrases (e.g., "Ignore all previous instructions") to explain how the platform moderates content. It is documentation of security features, not an injection attempt.
[COMMAND_EXECUTION]: The Python script hooks/scripts/agentscript-syntax-validator.py uses subprocess.run to execute standard Salesforce CLI (sf) commands. This is a legitimate functional requirement to validate the existence and status of Salesforce metadata (Apex classes, Flows) referenced in the user's Agent Script against their connected org. The script uses basic sanitization (_sql_quote) when constructing SOQL queries from the parsed script content.
[DATA_EXPOSURE]: No hardcoded credentials, unauthorized network calls, or sensitive file path access were detected. Configuration examples use standard placeholders.
[OBFUSCATION]: No evidence of Base64, hex-encoded payloads, homoglyph attacks, or hidden characters was found in any of the files.

sf-ai-agentscript