plan-decompose

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION] (MEDIUM): The skill dynamically assembles shell commands using the 'uv run' utility. Specifically, it executes a local script '.claude/skills/github-tools/scripts/gh.py' with arguments such as '--title' and '--body' which are directly populated with text parsed from 'plan.md'. Without explicit sanitization or shell-escaping, this could allow for argument injection (e.g., if a work unit title starts with '--' or contains shell metacharacters).
  • [PROMPT_INJECTION] (LOW): The skill is vulnerable to Indirect Prompt Injection (Category 8). Evidence Chain: 1. Ingestion points: The skill reads work unit details from 'plan.md'. 2. Boundary markers: None. There are no instructions for the agent to use delimiters or ignore instructions within the plan content. 3. Capability inventory: Execution of subprocesses ('uv run') and GitHub API interactions. 4. Sanitization: Absent. The skill instructions do not specify any validation or filtering of the content extracted from the plan file before it is used to generate tool commands.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:39 PM