memory-sync-setup
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a local bash script
scripts/generate.shto orchestrate project-wide operations including scanning the repository structure, emitting the generated skill files, and validating the installation status. - [INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting and processing instruction memory files (e.g., CLAUDE.md, .claude/rules/**), which creates an attack surface for indirect prompt injection. To mitigate this, the skill's design includes a Phase 3 'VERIFY' stage where independent read-only checker agents validate all proposed edits against the actual repository source code before they are finalized.
- [DYNAMIC_EXECUTION]: The
generate.shutility script uses theevalcommand for internal shell variable indirection and to execute predefined file search commands during the project enumeration phase. These execution paths are controlled within the script and do not process external untrusted input as code. - [DATA_EXPOSURE]: The generator script accesses the platform's project-specific metadata directory (
~/.claude/projects/) to resolve the project's memory configuration. The skill explicitly excludes sensitive directories such as secrets, build outputs, and git-ignored scratch files from its processing scope.
Audit Metadata