fix-php-vuln
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from external git repositories, Dockerfiles, and configuration files, which constitutes a vulnerability surface for indirect prompt injection attacks.
- Ingestion points: Git repositories, Dockerfiles, Composer dependency files, CI/CD pipeline configurations, and PHP source code analyzed during the automated discovery phase in SKILL.md.
- Boundary markers: The instructions do not provide explicit delimiters or "ignore embedded instructions" warnings for the agent when processing external file content.
- Capability inventory: The skill leverages powerful capabilities including shell commands (ripgrep, git), Docker image builds and pulls, and the execution of the PHP toolchain (Composer, PHPUnit, Rector, PHPCompatibility) to analyze and verify targets.
- Sanitization: There are no explicit instructions for the agent to sanitize or escape data harvested from target repositories before interpolation into prompts.
- Mitigation: The architecture implements an explicit approval gate (Step 4) that halts the execution flow and requires human review of the inventory and patching plan before any writes or deployments occur.
- [COMMAND_EXECUTION]: The skill directs the agent to utilize standard shell-based development and virtualization tools to perform its inventory and testing tasks.
- Evidence: Instructions include the use of
git cloneandrgfor repository analysis,docker buildanddocker pullfor artifact creation and verification, and various PHP ecosystem tools for regression testing and compatibility checks.
Audit Metadata