The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like git log to verify file changes and property routes. It also mandates running test suites provided by the TESTER role against code from the DEV role. While this is the intended functionality for a verification agent, executing code from various internal contributors carries inherent risk.
[INDIRECT_PROMPT_INJECTION]: The skill processes and executes untrusted data in the form of code and tests provided by other agents or users. This creates a vulnerability where a malicious entity could inject commands into the test suite or application logic that the LEAD agent will then execute with its own permissions.
Ingestion points: Files in the work branch, specifically code from the DEV role and test files from the TESTER role.
Boundary markers: None identified. The skill trusts the test runner to execute the provided tests.
Capability inventory: Execution of shell commands via git log and the execution of test runners; file writing to informes/ directory.
Sanitization: No specific sanitization or sandboxing of the test execution environment is mentioned in the instructions.

veredicto