framework-migration-deps-upgrade
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructions and provided code patterns involve ingesting external data from package changelogs and software registries to generate migration guides and risk assessments. This introduces a surface for indirect prompt injection if a malicious dependency version contains instructions intended to manipulate the agent's output or execution steps.\n
- Ingestion points: External changelog data and package metadata are fetched and parsed in the BreakingChangeDetector class within resources/implementation-playbook.md.\n
- Boundary markers: The generated migration guides and priority matrices do not use clear delimiters or negative constraints to prevent the agent from following instructions potentially embedded in the fetched changelog text.\n
- Capability inventory: The skill is designed to plan and execute system commands including npm install, git, and framework-specific CLI tools (e.g., ng, npx), as demonstrated in resources/implementation-playbook.md.\n
- Sanitization: There is no evidence of sanitization, filtering, or validation of external text content before it is interpolated into reports or used for decision-making.\n- [COMMAND_EXECUTION]: The skill provides code snippets and instructions that execute shell commands and system binaries to perform dependency analysis and updates as part of its primary function.\n
- Evidence: resources/implementation-playbook.md contains multiple instances of subprocess.run calls for npm and pip, as well as instructions for executing git, ng, and react-codemod commands.
Audit Metadata