The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill analyzes external Apex source code, which serves as an ingestion point for untrusted data. Maliciously crafted comments or strings within the code under test could potentially influence the agent's logic during test generation or rubric scoring.\n
Ingestion points: Phase 1 fetches class and trigger bodies from the Salesforce org or local repository.\n
Capability inventory: The skill can write and update metadata in the Salesforce org using tools such as metadata_create and metadata_update.\n
Boundary markers: There are no explicit instructions for the agent to use delimiters or sanitization when processing the fetched code content.\n- [DYNAMIC_EXECUTION]: The skill's core functionality involves generating new Apex test classes at runtime and deploying them to the Salesforce environment. While this is the intended purpose for a test generation tool, it represents the runtime creation and execution of code within the targeted organization.\n- [COMMAND_EXECUTION]: The skill uses local shell commands (e.g., test -f sfdx-project.json, command -v sf, sf org display) to detect the environment mode and verify the presence of the Salesforce CLI and Python. These are standard system discovery commands used for configuration and environment verification.\n- [EXTERNAL_DOWNLOADS]: The skill handles large data payloads by downloading artifacts from URLs provided in MCP tool responses (specifically the artifactUrl field). This mechanism is used for handling pagination and large metadata files from the connected Salesforce environment.

sf-test