product-docs-sync
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill explicitly instructs the agent to treat external content (issue bodies, PR descriptions, comments, and diffs) as data to analyze, not as instructions to follow, which is a key defense against prompt injection.
- [DATA_EXFILTRATION]: No network operations, API calls, or remote data transfer patterns were identified. The skill is instructed not to call GitHub APIs and to work only with local context files.
- [COMMAND_EXECUTION]: The skill does not contain any shell commands, subprocess calls, or privilege escalation patterns. It explicitly prohibits Git operations (staging, committing, or pushing) or modifying non-documentation files.
- [CREDENTIALS_UNSAFE]: No hardcoded secrets, environment variable access, or sensitive credential path access (e.g., .ssh, .aws) were found.
- [INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for untrusted data (PRs and Issues) but includes mandatory boundary instructions and strict capability scoping (only writing to documentation paths and a specific JSON result file), effectively mitigating the risk.
Audit Metadata