task-decomposition
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFENO_CODE
Full Analysis
- [NO_CODE]: The skill consists entirely of Markdown instructions and templates. There are no scripts (Python, JavaScript, shell) or binaries included in the package.
- [PROMPT_INJECTION]: I analyzed the instructions for any attempts to bypass safety filters or override system instructions. The content is strictly focused on task decomposition logic (MECE heuristics, diagnostic states, output modes) and contains no adversarial injection patterns.
- [EXTERNAL_DOWNLOADS]: There are no URLs, network requests (curl, wget), or external package dependencies identified in the skill files.
- [DATA_EXFILTRATION]: The skill does not access sensitive files, environment variables, or hardcoded credentials. It operates solely on the input provided by the user during the conversation.
- [DYNAMIC_EXECUTION]: There are no patterns for runtime code generation or dynamic loading. The use of Mermaid graphs and ASCII trees is for visualization of text only.
Audit Metadata