The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from local component files which could contain malicious instructions designed to influence the agent's code generation process.
Ingestion points: In Phase 1 and 2, the agent reads component source files (based on the $ARGUMENTS path) and existing .spec.ts files.
Boundary markers: There are no explicit boundary markers or instructions telling the agent to ignore prompt-like content within the code it analyzes.
Capability inventory: The skill has the capability to read files, write files, and execute shell commands (yarn, nx).
Sanitization: The skill lacks sanitization routines for the content extracted from the component files before it is interpolated into the test generation logic.
[COMMAND_EXECUTION]: The skill triggers the execution of shell commands to format and run generated code.
Evidence: Phase 4 explicitly instructs the agent to run yarn format and nx run <library>:test. This workflow involves the execution of agent-generated code, which is a potential risk if the generation was influenced by an indirect prompt injection.

create-test