banana-skill-finder
Warn
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONCREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTIONPERSISTENCEEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the installation of external code using
npx skills addandnpx add-skill. These commands are used to download and install packages from arbitrary GitHub repositories discovered through automated search results. Because the skill triggers proactively and suggests one-command installation of unvetted third-party software, it poses a risk of executing malicious code if search results are compromised or lead to untrusted repositories. - [CREDENTIALS_UNSAFE]: The skill manages a
SKILLSMP_API_KEYby reading it from environment variables and sending it to theskillsmp.comdomain viacurl. Furthermore, the installation instructions guide users to store this sensitive API key in plaintext within shell profile files (~/.bashrc,~/.zshrc) or local.envfiles, which are not secure credential storage mechanisms. - [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from multiple external sources to drive its proactive recommendation logic.
- Ingestion points: User search queries and responses from SkillsMP, skills.sh, and GitHub APIs (SKILL.md, references/skill_sources.md).
- Boundary markers: None; search results and user inputs are processed and interpolated into installation recommendations without delimiters or instructions to ignore embedded content.
- Capability inventory: The skill has the capability to execute shell commands via
npx, perform network operations viacurl, and enumerate local directory structures vials(SKILL.md). - Sanitization: There is no evidence of filtering, validation, or sanitization of the data retrieved from external skill registries before it is presented to the user as a recommended action.
- [PERSISTENCE]: The skill provides explicit instructions for users to modify their shell startup configuration files to export environment variables. This creates a persistence mechanism that ensures credentials and configurations are automatically loaded in every new session.
- [EXTERNAL_DOWNLOADS]: The skill makes network requests to
skillsmp.com,skills.sh, andapi.github.comfor search operations and triggers the download of external code via thenpxpackage runner.
Audit Metadata