Skills
Navigate
skills/calionauta/stelow/stelow-workflow-shape-up/Gen Agent Trust Hub

stelow-workflow-shape-up

Fail

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: Documentation in references/cli-tools/cymbal.md recommends installing a third-party tool using a risky pattern: irm https://raw.githubusercontent.com/1broseidon/cymbal/main/install.ps1 | iex. This command downloads and executes a PowerShell script from an untrusted GitHub repository without prior validation.\n- [DYNAMIC_EXECUTION]: The skill performs dynamic loading of code in SKILL.md by using the source command on a path computed at runtime (../../stelow-workflow-orchestrator/references/cli-tools/read-config.sh). It also uses node -e to execute JavaScript code generated within the shell script to parse configuration files.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from stelow.json and codebase analysis tools.\n
  • Ingestion points: stelow.json (workflow names) and codebase metadata via context/tech-preview.md.\n
  • Boundary markers: Absent; untrusted data is interpolated directly into the shaping process.\n
  • Capability inventory: File system writes to the .stelow/ directory and shell command execution (bash, node).\n
  • Sanitization: Absent; ingested strings are used without validation in shell contexts and generated proposals.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Sep 9, 2026, 12:37 AM