The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes test result output from the Salesforce CLI, which originates from an external environment (the Salesforce org being tested). This data is parsed and used to generate 'Agentic Fix Instructions' for the AI agent, creating a potential surface for indirect prompt injection.
Ingestion points: The hooks/scripts/parse-test-results.py script (line 335) reads the TOOL_OUTPUT environment variable, which contains the stdout from sf apex run test and sf apex get test commands.
Boundary markers: The formatted output provided to the agent lacks explicit delimiters or instructions to ignore potential directives embedded within the error messages or stack traces.
Capability inventory: The skill facilitates shell command execution via the sf CLI and delegates code modification tasks to the sf-apex skill, which allows the agent to write and deploy source code based on the parsed test failures.
Sanitization: The failure analysis script truncates error messages to 200 characters and uses regular expressions to extract specific error patterns, but it does not perform sanitization to detect or neutralize prompt injection attempts hidden in logs.
[COMMAND_EXECUTION]: The skill is designed to execute Salesforce CLI commands such as sf apex run test, sf apex get test, and sf apex list log. These are standard tools for the skill's intended purpose of managing Salesforce testing workflows.
[DYNAMIC_EXECUTION]: The skill provides several Apex code templates in the assets/ directory (e.g., basic-test.cls, dml-mock.cls, test-data-factory.cls). The agent uses these templates to generate and deploy new test classes to the Salesforce environment as part of the automated testing and fix process.

sf-testing