repository-quality-gate-fixer
Pass
Audited by Gen Agent Trust Hub on May 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local git and project-specific verification commands (e.g., lint, test, build) to audit and repair repositories. These operations are scoped to the local environment and require explicit authorization for non-read-only tasks.\n- [EXTERNAL_DOWNLOADS]: The skill manages dependency installation using standard developer tools such as npm, pip, and cargo. It follows project-level documentation and requires user confirmation for global changes.\n- [DATA_EXFILTRATION]: The skill supports pushing code and monitoring remote CI results. These actions are intended for repository management and are governed by user-authorized modes (e.g., 'push' mode).\n- [PROMPT_INJECTION]: The skill mitigates indirect prompt injection risks by processing repository data under a strict priority model.\n
- Ingestion points: Processes all local repository files (e.g., AGENTS.md, README) and external signals like PR comments or CodeRabbit reviews.\n
- Boundary markers: Employs a 'Instruction Priority Model' where core safety rules and explicit user requests override instructions found in processed files.\n
- Capability inventory: Performs file editing and shell execution using standard development toolchains.\n
- Sanitization: The
repo_quality_probe.pyscript includes dedicated logic to redact API tokens, private keys, and other secrets from its output to prevent accidental exposure.
Audit Metadata