memory-management

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • EXTERNAL_DOWNLOADS (MEDIUM): The skill frequently invokes npx @claude-flow/cli. This command downloads the package from the npm registry and executes it at runtime. The package @claude-flow/cli does not originate from a trusted organization or repository as defined in the security guidelines.
  • COMMAND_EXECUTION (MEDIUM): The skill uses shell scripts (memory-backup.sh and memory-consolidate.sh) to execute system commands and the external CLI. This involves running executable logic outside of the core agent environment.
  • PROMPT_INJECTION (LOW): (Category 8
  • Indirect Prompt Injection) The memory management system provides a surface for indirect prompt injection attacks.
  • Ingestion points: Untrusted data enters the agent context via the memory store command in SKILL.md.
  • Boundary markers: There are no specified delimiters or instructions to the agent to ignore embedded commands within the retrieved memory.
  • Capability inventory: The skill possesses the ability to retrieve and display stored content to the agent via memory search and memory get commands.
  • Sanitization: No evidence of sanitization, filtering, or validation is present for the data being stored in or retrieved from the memory namespaces.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:06 PM