skillify
Warn
Audited by Gen Agent Trust Hub on May 3, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The preamble executes numerous local binaries found in
~/.claude/skills/gstack/bin/. It specifically usesevalandsourceto execute the output of scripts likegstack-slugandgstack-repo-mode, which is a high-risk pattern for dynamic configuration. - [REMOTE_CODE_EXECUTION]: The skill's primary purpose is to synthesize TypeScript code (
script.tsandscript.test.ts) from previous conversation history. It then executes this generated code usingbun test. This represents dynamic code generation and execution based on potentially untrusted historical input (scraped website data). - [DATA_EXFILTRATION]: The skill implements a telemetry system that logs skill usage, durations, and repository names to
~/.gstack/analytics/. Additionally, the 'GBrain Sync' feature is designed to publish session memory to a private GitHub repository. Although these features are gated behind anAskUserQuestionprompt, they represent the transmission of project context and session history to external locations. - [COMMAND_EXECUTION]: The skill performs git operations, including
git addandgit commit, to modify the project'sCLAUDE.mdfile to inject 'skill routing' rules. - [SAFE]: References to the external domain
garryslist.orgare treated as vendor-controlled resources belonging to the skill author.
Audit Metadata