qemu-agent-feedback
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEDATA_EXFILTRATIONDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXPOSURE]: The skill reads session logs, audit records, and command history from the
.oh-my-qemu/directory to construct feedback reports. This involves access to detailed records of previous agent activities. - [DATA_EXFILTRATION]: The skill uses the
ghCLI to submit issues to GitHub. This risk is mitigated by mandatory sanitization steps and a requirement for the user to approve the final content and target repository before submission. - [COMMAND_EXECUTION]: The skill executes the GitHub CLI (
gh) and a local bash script (check-redaction.sh). These operations are used for scanning output for private data and interacting with the authorized repository. - [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from local log files and command results.
- Ingestion points: Audit logs, command outputs, and task metadata stored in
.oh-my-qemu/. - Boundary markers: The skill uses a structured markdown template and a sanitization table, though it lacks explicit instructions to ignore embedded commands within the processed logs.
- Capability inventory: File system read/write, network access via
ghCLI. - Sanitization: A dedicated script (
check-redaction.sh) is used to verify that no sensitive data (tokens, paths, identities) remains in the draft before it is presented to the user.
Audit Metadata