report-bug-ce
Pass
Audited by Gen Agent Trust Hub on Mar 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: Transmits user-provided bug reports and system environment information (OS info via
uname, CLI versions, and plugin metadata) to the external GitHub repositoryEveryInc/compound-engineering-plugin. This operation is the primary purpose of the skill and targets a repository owned by the skill's author. - [COMMAND_EXECUTION]: Utilizes shell commands including
uname -a,gh issue create, and CLI version flags (e.g.,claude --version) to collect metadata and perform the bug submission. - [PROMPT_INJECTION]: Presents an indirect prompt injection surface where untrusted user input (bug details) is interpolated into a formatted bug report and a shell command (
gh). - Ingestion points: User responses to questions regarding bug categories, behavior, and steps to reproduce in
SKILL.md. - Boundary markers: Absent.
- Capability inventory: Execution of
gh issue create(network/write) inSKILL.md. - Sanitization: No specific instructions are provided to sanitize or escape user-supplied strings before they are included in the shell command or the report body.
Audit Metadata