go-workspace-skill
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/workspace.pyutilizes thesubprocessmodule to executegitfor repository synchronization andgofor building and testing Go projects. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository metadata from a configuration file, which presents an attack surface where malicious data could influence agent operations.
- Ingestion points: Repository names and URLs are ingested from a
.go-workspace.jsonconfiguration file, as handled inscripts/workspace.py. - Boundary markers: The skill enforces a strict JSON schema and verifies the existence of required fields during the configuration loading phase.
- Capability inventory: The tool possesses the capability to clone remote repositories, execute localized build/test commands, and perform recursive directory deletion via
shutil.rmtree. - Sanitization: Repository names undergo strict validation in the
invalid_name_reasonfunction, which prevents path traversal by rejecting names containing slashes, backslashes, colons, or '..' patterns, and ensuring names are not absolute paths.
Audit Metadata