create
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
Bashtool to perform various file system and repository management tasks, such as creating directory structures (mkdir), modifying configuration files (jq,Edit), and executing build and release workflows (npm run release,moon run repo:release-full). - [INDIRECT_PROMPT_INJECTION]: The skill accepts input for the plugin name and category through user prompts and command arguments, which are then used directly in shell commands and configuration files.
- Ingestion points: User input from
AskUserQuestionand theargument-hintparameter inSKILL.md. - Boundary markers: No delimiters or warnings are used when interpolating these variables into bash scripts.
- Capability inventory:
Bash,Write,Edit,Skill,Task. - Sanitization: There is no evidence of validation or escaping for the
$PLUGIN_NAMEvariable before it is used in path construction and command execution. - [CREDENTIALS_UNSAFE]: In
references/phase4-release.md, the skill retrieves a GitHub token usinggh auth tokenand assigns it to an environment variable (GITHUB_TOKEN) for use by thenpmrelease process. This involves the handling of sensitive credentials within the agent's task environment.
Audit Metadata