ios-fix
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute local utility binaries located in
~/.claude/skills/gstack/bin/. These utilities handle configuration, telemetry, and project state. It also invokes standard development tools likexcodebuildandswift runto compile and deploy iOS applications. - [DATA_EXFILTRATION]: The skill implements a telemetry system that logs usage data (skill name, duration, outcome, and repository name) to local files and potentially transmits it to the vendor's infrastructure via
gstack-telemetry-log. This behavior is transparently disclosed in the preamble and is part of the tool's analytics. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. In Phase 1 and 2, it ingests external data from bug reports (
/ios-qafindings) and Swift source code. There are no explicit instructions for sanitizing this data or using boundary markers before the agent processes it to determine the root cause of a bug. - [COMMAND_EXECUTION]: The preamble uses
evalandsourceon the output of local binaries (gstack-slugandgstack-repo-mode). This involves dynamic execution of content generated at runtime, which is a standard pattern for the gstack infrastructure but relies on the integrity of those local binaries. - [COMMAND_EXECUTION]: The skill automates the creation and modification of project files, specifically appending 'Skill routing' rules to
CLAUDE.md. This is a persistence mechanism for the agent's behavior within a specific project context.
Audit Metadata