The Agent Skills Directory

[SAFE]: The skill instructions focus on local diagnostics and formatting. No patterns of data exfiltration, credential theft, or unauthorized network communication were detected.
[COMMAND_EXECUTION]: The skill uses local commands such as ptop for code formatting and command --version for tool detection. These commands are executed within the context of the local project directory and are restricted to standard development utilities.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source files (.pas, .dpr, .dpk, .lpr) and project configurations. This represents a potential vulnerability surface, which is handled with procedural safeguards.
Ingestion points: Pascal and Delphi source code files, as well as existing repository linting configurations (SKILL.md).
Boundary markers: Not explicitly defined in the instructions, but the skill workflow requires identifying the specific scope before execution.
Capability inventory: Command execution capabilities are limited to local formatting tools and version checks across all scripts.
Sanitization: The skill relies on manual user intervention, requiring the review of diffs and explicit confirmation before any modifications are committed.

codeguard-pascal