The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external files or modules provided through the $ARGUMENTS parameter. This creates a surface for indirect prompt injection if the target codebase contains malicious instructions or data meant to influence the agent's behavior during analysis or test execution.
Ingestion points: The target file or module path specified in $ARGUMENTS within SKILL.md.
Boundary markers: The skill does not define clear boundaries or include instructions to ignore embedded commands within the processed data.
Capability inventory: The skill instructs the agent to "Generate coverage report," "Write tests," and "Run full test suite," which involve file system interaction and code execution via the agent's tools.
Sanitization: There is no evidence of sanitization or validation logic to ensure the target files do not contain harmful payloads.
[PROMPT_INJECTION]: The final instruction in SKILL.md uses an "IMPORTANT" directive to command the agent to "automatically exit" after completion. This is an attempt to override the agent's standard control flow and session management.

improve-coverage