Skills
Navigate
skills/open-gsd/gsd-core/gsd-ultraplan-phase/Gen Agent Trust Hub

gsd-ultraplan-phase

Pass

Audited by Gen Agent Trust Hub on Oct 1, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill accepts user input through the $ARGUMENTS variable, which is processed by the ultraplan-phase workflow. This creates a surface for indirect prompt injection where untrusted input could influence agent behavior.
  • Ingestion points: $ARGUMENTS field in SKILL.md.
  • Boundary markers: None present in the skill definition to isolate user arguments from instructions.
  • Capability inventory: The skill is authorized to use Bash, Read, Glob, and Grep tools to execute its workflow.
  • Sanitization: No evidence of input validation or sanitization is present in the skill instructions.
  • [EXTERNAL_DOWNLOADS]: The skill is designed to offload data to the ultraplan cloud infrastructure associated with the claude.ai service. This involves transmitting local repository context to a remote service for processing.
  • [COMMAND_EXECUTION]: The skill executes a local workflow (~/.claude/gsd-core/workflows/ultraplan-phase.md) that utilizes the Bash tool to coordinate between the local environment and the remote cloud session.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 1, 2026, 07:49 PM