graphicode-start-ts-react
Pass
Audited by Gen Agent Trust Hub on Apr 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local shell commands (
cat ./graphig.md,ls -d) to read project metadata and determine the directory structure for code generation. - [PROMPT_INJECTION]: Indirect injection surface identified: the skill reads untrusted data from local configuration files (
graphig.mdandstate.graphig.md) and uses this information to generate application code (launcher.ts). - Ingestion points: Reads metadata from
./graphig.mdand description fields instate.graphig.mdacross multiple directories. - Boundary markers: Absent; the skill trusts the content of these metadata files for code generation.
- Capability inventory: Writing new source files (
launcher.ts,playground.ts), modifying build configurations (vite.config.ts,webpack.config.ts), and updatingpackage.jsonscripts. - Sanitization: None detected; the agent is instructed to use identifiers and descriptions found in the files directly in the output code.
Audit Metadata