memory-init
Pass
Audited by Gen Agent Trust Hub on Mar 6, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill injects mandatory behavioral directives into the agent's project-level instruction file (CLAUDE.md).
- Evidence: Step 5c defines instructions like 'MANDATORY pre-response search' and 'No exceptions — always search first, respond second' which are designed to override the agent's default decision-making logic.
- [COMMAND_EXECUTION]: The skill performs extensive file system operations to set up the memory environment.
- Evidence: The workflow creates and populates files including .claude/memory-settings.json, .claude/memory_counter.txt, and .claude/memory-reminder.md.
- Evidence: The skill reads template files from the local environment via the ${CLAUDE_PLUGIN_ROOT} variable and writes them to the project root.
Audit Metadata