Two tools with clear security-focused purpose, but significant gaps in parameter documentation and output schema specification. Both tools have descriptions (good), but lack detailed parameter descriptions and explicit output schemas. The audit_workflow tool accepts a single string parameter with minimal guidance on expected format or constraints. check_expression_injection similarly lacks parameter validation hints. No error handling guidance visible. Naming is verb-forward and clear (audit_*, check_*), following pattern:tool conventions. No evidence of output schema documentation.

• Document the output schema for audit_workflow: specify the structure of findings (e.g., array of {issue_type, severity, line_number, description, remediation}). This enables LLMs to extract and act on results.
• Expand parameter descriptions: 'workflow_yaml' should note 'Complete GitHub Actions workflow YAML file (2-space indentation, standard GitHub Actions syntax). Max 100KB.' Add format constraints and size limits.
• Add error handling guidance: 'If YAML is invalid, returns {error: string, line_number: int}. If workflow is valid but empty, returns {findings: []}.' This tells LLMs what to expect on failure.
• Clarify tool selection: update check_expression_injection description to say 'Use this for a single shell command; use audit_workflow to check an entire workflow file.' Prevents LLM confusion.
• Document output limits: 'audit_workflow returns up to 50 findings per call, sorted by severity. If more exist, include a truncation notice.' Prevents context window exhaustion.