The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted code files (C# and Python), creating a theoretical attack surface for indirect prompt injection. However, it implements robust mitigations, including strict boundary markers defined in the 'Freeze the relevant contract' section and a formal candidate verification lifecycle to ensure equivalence. Ingestion points include the authorized implementation and tests (SKILL.md). Capabilities include file modification and command execution for verification. The skill explicitly instructs the agent to redact secrets and sensitive values from all reports to prevent accidental data exposure.
[COMMAND_EXECUTION]: The skill uses standard, repository-native development tools for verification, such as dotnet test, pytest, mypy, and ruff. These commands are restricted to non-restoring modes (e.g., --no-restore) and are only used when already present in the user's environment to verify behavior equivalence.
[EXTERNAL_DOWNLOADS]: The skill references official documentation from trusted sources (Microsoft and the Python Software Foundation). It includes explicit prohibitions against installing new analyzers, packages, or runtimes, directing all proposed dependency changes to a dedicated auditing skill.

reduce-code-slop