sync-public-skills
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [COMMAND_EXECUTION]: The skill's primary workflow involves directing the agent to run PowerShell scripts, such as
Sync-AndPush.ps1andCommit-DirtySkills.ps1, and Git operations. These commands are used to automate repository commits, state management, and synchronization tasks. - [INDIRECT_PROMPT_INJECTION]: The agent ingests and processes content from external directories, including
.github/skills/and thecopilot-skills/inventory. Because this data influences the synchronization and auditing logic, malicious instructions embedded in these files could theoretically attempt to manipulate the agent's behavior. - Ingestion points: Files located in
.github/skills/andcopilot-skills/within the private repository. - Boundary markers: The skill documentation specifies several "Gates" (New Skill Classification, Agent Discovery, Copilot-Skills Private Inventory) designed to prevent unauthorized synchronization.
- Capability inventory: The skill utilizes Git commands, shell script execution (PowerShell), and file reading/writing for placeholder generation and sensitive data masking.
- Sanitization: The skill includes instructions for "sensitive scans" and manual classification requirements for unknown skills before synchronization can proceed.
- [DATA_EXFILTRATION]: The skill is designed to move content from private source repositories to public or internal destinations. While it provides detailed instructions for masking secrets, using placeholders, and performing sensitive data removal, the tool's core functionality is the external transfer of data.
Audit Metadata