plankton-code-quality

Warn

Audited by Gen Agent Trust Hub on Oct 14, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill documentation describes a complex system of PreToolUse and PostToolUse hooks that execute shell scripts (multi_linter.sh, protect_linter_configs.sh, stop_config_guardian.sh) and various CLI tools (ruff, biome, uv, bun, jaq, etc.) on every file edit. It also implements command interception to block legacy package managers like pip and npm in favor of uv and bun.\n- [DYNAMIC_EXECUTION]: The skill describes a Phase 3 where it autonomously spawns 'claude -p' subprocesses. These sub-agents are routed to different model tiers (Haiku, Sonnet, Opus) based on linter violations and are tasked with modifying the codebase without direct user oversight.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data in the form of linter violation reports (JSON) and feeds this content into the prompt of a sub-agent to drive automated code fixes. This represents an attack surface where maliciously crafted code that triggers specific linter messages could influence the agent's behavior.\n
  • Ingestion points: JSON violation data from linting tools processed in multi_linter.sh.\n
  • Boundary markers: None described in the prompt interpolation logic for sub-agents.\n
  • Capability inventory: Spawning subprocesses, shell command execution, and file writing.\n
  • Sanitization: None described for the violation messages passed to the sub-agent.\n- [EXTERNAL_DOWNLOADS]: The documentation points users to an external repository for installation and lists numerous third-party dependencies across multiple languages (Python, Node.js, Shell, etc.) including ruff, uv, biome, and semgrep.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Oct 14, 2026, 02:58 PM