skill-management

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): The skill documents and enables the use of dynamic context injection using the backtick syntax (represented as ❗️command in the text). This allows for the execution of arbitrary shell commands on the host system to populate prompt context, which is a high-privilege capability.- REMOTE_CODE_EXECUTION (MEDIUM): The support for $ARGUMENTS interpolation into dynamic shell commands (e.g., in the 'pr-summary' example) creates a direct path for shell injection if arguments are not properly sanitized by the calling agent.- INDIRECT_PROMPT_INJECTION (LOW): The skill is intended to 'Create, update, and manage' other skills, creating an attack surface for persisting malicious instructions.
  • Ingestion points: The skill logic processes user/agent input to write and modify SKILL.md files in ~/.claude/skills/ or .claude/skills/.
  • Boundary markers: The documentation lacks instructions for implementing delimiters or 'ignore' instructions when handling untrusted user input.
  • Capability inventory: The system supports shell command execution (!command) and subagent forking (context: fork).
  • Sanitization: No sanitization or validation logic is specified for the input stored in managed skill files.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:42 PM