writing-claude-directives
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists entirely of instructional markdown and Graphviz documentation. It contains no executable scripts, automated commands, or tool definitions that could interact with the host system.
- [EXTERNAL_DOWNLOADS]: The
long-running-state-patterns.mdfile contains references to official Anthropic engineering blog posts and two third-party GitHub repositories (ruvnet/claude-flow,AnandChowdhary/continuous-claude) provided as learning resources for orchestration and CI/CD patterns. These are informational links and do not involve automated code execution. - [PROMPT_INJECTION]: While the skill provides templates for modifying agent behavior (e.g., 'Action Bias' templates), these are presented as design patterns for developers to incorporate into their own system prompts. The skill explicitly emphasizes security awareness by requiring the use of a security hardening skill when handling secrets.
Audit Metadata