lov-skill-publisher

Pass

Audited by Gen Agent Trust Hub on Aug 27, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands for version control and publishing, including git, gh (GitHub CLI), and python3. It also uses curl to send revalidation requests to a configured endpoint.\n- [DATA_EXFILTRATION]: The skill reads from a local profile directory ($HOME/.skill-publisher/) to manage user settings and utilizes environment variables for secrets, such as LOVSTUDIO_REVALIDATE_SECRET, which is passed in HTTP headers during revalidation to potentially external endpoints.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes local Skill source files and metadata (SKILL.md, connector-meta.json) which could potentially contain malicious instructions intended to mislead the agent during the build or publishing process.\n
  • Ingestion points: Local skill files and connector-meta.json processed in scripts/build_workbuddy.py and scripts/validate_skill.py.\n
  • Boundary markers: Absent in the processing scripts.\n
  • Capability inventory: Full shell command execution and file system access via Python standard libraries.\n
  • Sanitization: Limited to metadata field validation in the Python scripts.\n- [EXTERNAL_DOWNLOADS]: The skill communicates with external platforms including GitHub, codebuddy.cn, and skillpay.alipay.com for package submission and verification. These are established services consistent with the skill's stated purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 27, 2026, 06:58 AM
Security Audit — agent-trust-hub — lov-skill-publisher