skills/wei18/upkeep/upkeep-audit/Gen Agent Trust Hub

upkeep-audit

Pass

Audited by Gen Agent Trust Hub on Jul 29, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches the Upkeep auditing tool directly from the author's GitHub repository at https://github.com/wei18/upkeep during the setup process.
  • [REMOTE_CODE_EXECUTION]: Executes a shell script (local-audit.sh) and installs Node.js packages using npm ci from the downloaded repository content.
  • [COMMAND_EXECUTION]: Uses local shell commands to manage the auditing environment, including git clone, git pull, and the execution of parallel audit reviewers via the claude CLI.
  • [PROMPT_INJECTION]: The skill processes untrusted local repositories which presents a surface for indirect prompt injection.
  • Ingestion points: Reads files and metadata from the user-specified <target-path> repository.
  • Boundary markers: None identified; the instructions do not specify delimiters for treating the target repository content as untrusted data.
  • Capability inventory: Subprocess execution (claude -p), file system read access, and script execution via local-audit.sh.
  • Sanitization: No explicit sanitization or filtering of the ingested repository content is mentioned in the skill instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 29, 2026, 02:47 AM