github-skill-forge

Warn

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/forge.py uses subprocess.run(..., shell=True) to execute git clone commands. While the input URL is validated against a regular expression, the use of a shell environment for dynamic commands is a security risk that can lead to command injection if validation is bypassed or if shell-specific characters like backticks are included in the input.
  • [INDIRECT_PROMPT_INJECTION]: The skill aggregates untrusted data (READMEs, code snippets, file trees) from external GitHub repositories into a context_bundle.md file, which is then used as primary context for the AI Agent. This creates a surface for indirect prompt injection attacks.
  • Ingestion points: scripts/forge.py downloads content from arbitrary repositories via the online_repo_scanner and clone_repository functions.
  • Boundary markers: The skill uses simple Markdown headers but lacks robust delimiters or instructions to the Agent to ignore embedded commands within the ingested data.
  • Capability inventory: The skill can execute shell commands and write files, which could be abused if an injection attack succeeds in influencing the Agent's decision-making process.
  • Sanitization: There is no evidence of sanitization or filtering of the content retrieved from remote repositories before it is processed by the Agent.
  • [REMOTE_CODE_EXECUTION]: The skill's primary function is to download and install third-party code from the internet and prepare it for execution as a skill. This workflow facilitates the execution of unverified remote code on the host system.
  • [DATA_EXFILTRATION]: The scripts/forge.py script sends API requests that include the user's GITHUB_TOKEN to several unofficial GitHub mirror and proxy domains (e.g., vps.sc, gitmirror.com, 99988866.xyz). Communicating sensitive access tokens to these untrusted third-party services poses a risk of credential exposure.
  • [EXTERNAL_DOWNLOADS]: The skill performs automated downloads from various third-party domains and mirrors that are not part of the official GitHub infrastructure or recognized trusted vendors.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 20, 2026, 08:26 AM