opencli-autofix
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external websites during the diagnostic phase, which could contain malicious instructions designed to influence the agent's code-patching behavior.
- Ingestion points: Reads trace files (summary.md, trace.jsonl, network.jsonl) and browser state in Step 1, 2, and 3.
- Boundary markers: The instructions lack explicit directives to treat data in trace artifacts as untrusted or to utilize delimiters when processing browser snapshots.
- Capability inventory: The skill possesses Edit and Write permissions to local files and uses the GitHub CLI tool to create issues.
- Sanitization: No specific sanitization or validation of the ingested web content is described before it is used for analysis and patching.
- [COMMAND_EXECUTION]: The skill relies on executing shell commands for diagnostics, browser exploration, and upstream reporting.
- Utilizes Bash for opencli commands to collect traces and inspect browser state.
- Utilizes Bash for gh issue create to submit reports to the jackwener/OpenCLI repository.
Audit Metadata