gstack-upgrade
Fail
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill executes the Bun installer via a pipe to bash when the local environment needs a version refresh. The source (bun.sh) is a well-known service, making this a standard installation procedure.
- Evidence:
curl -fsSL https://bun.sh/install | bashinSKILL.md. - [EXTERNAL_DOWNLOADS]: Fetches tool updates and browser dependencies from the author's GitHub repository and Playwright's official installation path. These resources originate from the trusted vendor (garrytan) or well-known services.
- Evidence:
git clone https://github.com/garrytan/gstack.gitinSKILL.mdandbunx playwright install chromiuminmigrations/v1.65.0.0.sh. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from local changelogs and git logs to generate a summary for the user, creating a potential surface for embedded instructions in project documentation to influence the agent's behavior.
- Ingestion points:
CHANGELOG.mdandgit logoutput inSKILL.mdStep 6. - Boundary markers: No explicit delimiters or ignore-instructions are used for the summary output.
- Capability inventory: Uses the
Bash,Read,Write, andAskUserQuestiontools. - Sanitization: Relies on the agent's summarization capabilities without programmatic pre-processing sanitization.
- [DYNAMIC_EXECUTION]: Migration scripts use Python one-liners as a fallback to resolve absolute file paths when system utilities are unavailable.
- Evidence: Inline Python
os.path.realpathcall inmigrations/v1.1.3.0.sh.
Recommendations
- HIGH: Downloads and executes remote code from: https://bun.sh/install - DO NOT USE without thorough review
Audit Metadata