skills/rheinmir/skills/fdk/Gen Agent Trust Hub

fdk

Fail

Audited by Gen Agent Trust Hub on Oct 14, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the agent to download and execute a shell script directly from a remote GitHub repository using process substitution. This 'curl piped to bash' pattern is a dangerous execution method that bypasses verification and security checks, potentially allowing for arbitrary code execution if the remote source or the transport is compromised.
  • Evidence: bash <(curl -fsSL https://raw.githubusercontent.com/Rheinmir/setup/orca/fdk/tools/fdk-kit.sh) pull found in the 'Adapt vào overstack remote' section of SKILL.md.
  • [EXTERNAL_DOWNLOADS]: The skill fetches resources from external URLs during its setup and update workflows, specifically targeting the author's own GitHub repository.
  • Evidence: https://raw.githubusercontent.com/Rheinmir/setup/orca/fdk/tools/fdk-kit.sh is used as a bootstrap script.
  • [METADATA_POISONING]: The skill includes explicit instructions (RULE-02) to remove all AI attribution, credits, or indicators from commits, code, and documentation. This is a deceptive practice that hides the provenance of the code from other developers and automated audits.
  • Evidence: RULE-02 (MUST): KHÔNG ghi công AI — commit message / PR / code / wiki KHÔNG được chèn Co-Authored-By: Claude… in SKILL.md.
  • [INDIRECT_PROMPT_INJECTION]: The skill takes arbitrary user-provided text ('mô tả của user') and interpolates it directly into a shell command for submission. Without explicit sanitization, this provides a surface for command injection attacks where a malicious user description could execute unintended shell commands.
  • Ingestion points: The 'Input và output contract' defines 'việc framework hoặc skill cần distill' as required user input.
  • Boundary markers: None. The input is passed directly into a shell command.
  • Capability inventory: The skill uses bash, python3, git, and custom kit scripts for file writing and network operations.
  • Sanitization: No sanitization or escaping logic is described for handling the user-provided description before it is passed to the fdk-kit.sh submit command.
  • [COMMAND_EXECUTION]: The skill performs extensive shell operations including repository cloning, running local Python scripts, and executing toolkit binaries which interact with the system and network environment.
  • Evidence: Instructions include python3 fdk/tools/ci-local.py, git clone, and gh pr create.
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
HIGH
Analyzed
Oct 14, 2026, 02:24 AM