The Agent Skills Directory

Indirect Prompt Injection (HIGH): The skill is designed to process external code provided via the $ARGUMENTS variable.
Ingestion Point: External source code provided for refactoring (SKILL.md).
Capability Inventory: The instructions 'Apply changes in small slices' and 'Update tests and verify regressions' imply the agent has file-system write access and the ability to execute shell commands (test runners).
Boundary Markers: There are no boundary markers or instructions to ignore embedded commands within the data being processed.
Sanitization: No sanitization or validation of the input code is performed.
Risk: An attacker could embed malicious instructions in comments or string literals within the code provided for refactoring, which could then be executed by the agent during the file-writing or testing phases.
Command Execution (HIGH): The instruction to 'Update tests' implies the execution of arbitrary test suites located in the user's environment.
Risk: If an indirect prompt injection succeeds, the agent could be tricked into running malicious scripts disguised as unit tests, leading to local code execution.
External Source (LOW): The skill originates from an untrusted third-party repository (github.com/sickn33/antigravity-awesome-skills), which has not been verified for safety or supply chain integrity.

refactor-clean