graphicode-init
Pass
Audited by Gen Agent Trust Hub on Mar 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local shell commands such as 'mkdir', 'cp', and 'cat' to create the project structure and copy internal assets. These operations are restricted to the local filesystem and are consistent with the skill's stated purpose.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by writing unvalidated user input (specifically the 'appName') into the 'graphig.md' file. This could allow malicious instructions to be stored in the configuration file and later interpreted by other skills or agents.
- Ingestion points: User-provided input for project information in SKILL.md (Step 1).
- Boundary markers: None present in the configuration file generation (Step 3).
- Capability inventory: Local file writing and directory creation via shell commands in SKILL.md.
- Sanitization: Input for 'language' and environments is validated against 'references/options.md', but 'appName' is not sanitized or escaped.
Audit Metadata