The Agent Skills Directory

[Indirect Prompt Injection Surface]: The skill is designed to ingest and analyze external content, specifically test method bodies and production source code provided by the user or referenced through file paths. This creates a potential surface where data could influence agent behavior if not properly isolated.
Ingestion points: Source code and test definitions are retrieved from the workspace or accepted as inline input during Step 2 of the workflow.
Boundary markers: While the instructions direct the agent to resolve specific code spans, they do not prescribe the use of explicit delimiters or instructions to ignore embedded content within the analyzed code.
Capability inventory: The skill is restricted to read-only analysis. It explicitly states that it does not perform file edits, command execution, or network operations (Workflow preamble).
Sanitization: The instructions do not define specific sanitization or filtering steps for the ingested code before analysis.
[Agent Tool Orchestration]: The skill coordinates with another tool, test-gap-analysis, to fulfill its objectives.
Evidence: The workflow includes instructions to "load test-gap-analysis by name once before scoring."
Context: This implementation represents standard orchestration within the agent platform. The instructions emphasize that this dependency is used strictly for read-only assessment, which is consistent with the skill's intended functionality.

grade-tests