The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection as it audits untrusted source code which could contain malicious instructions embedded in comments or strings.
Ingestion points: Processes project source files through the Read, Glob, and Grep tools (e.g., scanning src/ for client identifiers).
Boundary markers: The skill does not implement specific delimiters or safety instructions to distinguish between the skill's own logic and the content of the files being audited.
Capability inventory: The agent has access to Bash, Write, and Edit tools, which could be misused if an injection in the analyzed code successfully overrides instructions.
Sanitization: No sanitization or verification of code content is performed before processing.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform structural audits and codebase management.
Evidence: Uses grep and find to locate hardcoded client identifiers and misplaced files in the codebase.
Evidence: Suggests git mv and mkdir commands to reorganize files during refactoring operations.
Context: These operations are consistent with standard software engineering practices and are used within the defined scope of the skill's purpose.

fork-discipline