skills/rheinmir/setup/doyourmagic/Gen Agent Trust Hub

doyourmagic

Fail

Audited by Gen Agent Trust Hub on Sep 8, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTIONPERSISTENCE
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill uses dangerous patterns to execute remote code by downloading and piping shell scripts directly into a bash shell. Findings include: \n
  • bash <(curl -fsSL https://raw.githubusercontent.com/Rheinmir/setup/orca/fdk/tools/fdk-kit.sh) pull in SKILL.md.\n
  • curl -fsSL https://raw.githubusercontent.com/Rheinmir/setup/orca/harness/poc-vendor-neutral/bootstrap.sh | bash in dym-setup-install.skill.md.\n- [INDIRECT_PROMPT_INJECTION]: The skill has a significant attack surface for indirect prompt injection as it is designed to ingest and process external, untrusted repositories to generate runnable workflows. \n
  • Ingestion points: Reads external repository manifests (package.json, pyproject.toml), README files, and source code during the clone/explore phase.\n
  • Boundary markers: The instructions do not define boundary markers or delimiters to separate untrusted repo content from agent instructions.\n
  • Capability inventory: The skill has extensive capabilities including shell command execution, file system modification, and network access via curl.\n
  • Sanitization: No sanitization or validation of the ingested external content is mentioned before it is processed by the agent.\n- [PERSISTENCE]: The skill modifies global configuration files to maintain its influence across different agent sessions. \n
  • Modifies ~/.claude/settings.json to merge hooks and permissions in dym-setup-install.skill.md.\n
  • Installs global engine components in ~/.claude/harness/ and adds global skills via npx skills.\n- [DATA_EXFILTRATION]: The skill accesses and modifies sensitive configuration files in the home directory (~/.claude/settings.json) and performs network operations using curl. This combination provides a potential path for exfiltrating environment metadata or configuration secrets discovered during repo exploration.\n- [COMMAND_EXECUTION]: The skill executes multiple shell commands using arguments derived from external, untrusted repository content, such as manifest scripts and installer scripts.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/Rheinmir/setup/orca/fdk/tools/fdk-kit.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 8, 2026, 02:20 AM
Security Audit — agent-trust-hub — doyourmagic