qcsd-refinement-swarm

Warn

Audited by Gen Agent Trust Hub on Mar 11, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill uses npx @claude-flow/cli@latest in multiple phases (2, 4, 7). This command downloads and executes code from the NPM registry at runtime, which is a significant security risk if the package is compromised or malicious.\n- [COMMAND_EXECUTION]: Several CLI commands are executed to manage the swarm state, including swarm init, agent spawn, and memory store. These commands interact directly with the shell environment.\n- [EXTERNAL_DOWNLOADS]: The skill initiates downloads of external packages and potentially other resources via the npx and git clone (implied in CLI options) commands.\n- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it interpolates untrusted user story content into agent prompts with only basic delimiters.\n
  • Ingestion points: User story content processed in Phases 1, 2, and 4.\n
  • Boundary markers: Uses === STORY CONTENT START === and === STORY CONTENT END === delimiters.\n
  • Capability inventory: The skill can write files, execute shell commands via npx, and perform MCP memory operations.\n
  • Sanitization: No input validation or instruction filtering is applied to the story content before it is passed to sub-agents.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 11, 2026, 09:58 PM