The Agent Skills Directory

[PROMPT_INJECTION]: The skill analyzes commit diffs to recommend test suites, creating a surface for indirect prompt injection where malicious instructions could be embedded in the code changes being reviewed.
Ingestion points: Commit diffs are ingested and processed in references/recommendation-logic.md to classify change types and extract method names.
Boundary markers: The instructions do not define explicit delimiters or "ignore instructions" directives for the diff content provided to the agent.
Capability inventory: The skill possesses the capability to query Salesforce metadata and perform write operations on test suite assignments using the sf api request rest command, as detailed in references/api-endpoint.md.
Sanitization: No specific sanitization, validation, or escaping logic is described for the method names or change data extracted from the diffs.
Mitigation: This risk is significantly mitigated by the governance rules in SKILL.md and references/suite-assignment-modes.md, which require mandatory human-in-the-loop confirmation gates and impact previews (Mode C) before any API payload is executed.
[COMMAND_EXECUTION]: The skill uses sf data query and sf api request rest to interact with the Salesforce platform. These operations are restricted to standard objects (e.g., DevopsPipeline, DevopsTestSuiteStage) and official Connect API endpoints, following legitimate administrative patterns.
[EXTERNAL_DOWNLOADS]: The skill recommends the installation of the @salesforce/plugin-agent CLI plugin. As this plugin is an official tool from the same vendor as the skill author (forcedotcom/Salesforce), it is a trusted vendor resource and does not present a security risk.

dx-devops-test-suite-assignments-configure