deliver

Pass

Audited by Gen Agent Trust Hub on Nov 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill frequently executes system commands including git, gh, herdr, and orca via bundled Python scripts (deliver-runtime.py and deliver-orca-preflight.py). Analysis of these scripts shows they use subprocess with list arguments and regex-based input validation for identifiers (e.g., [a-zA-Z0-9_-]{1,80}), which effectively mitigates shell injection risks.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources such as GitHub issues, project specifications, and ticket trackers. This creates a surface for indirect prompt injection. The skill implements a critical security control by requiring a mandatory 'Plan Stop' where the agent must present its proposed implementation plan to the user for manual approval before any code is modified or executed.
  • [EXTERNAL_DOWNLOADS]: The skill interacts with remote repositories using git and the GitHub CLI (gh). These operations are restricted to the repository context and are necessary for the skill's primary purpose of software delivery.
  • [SAFE]: No malicious patterns such as hardcoded credentials, data exfiltration, or obfuscation were detected. The Python scripts demonstrate defensive coding practices, such as using atomic file writes, validating process ownership, and enforcing private state directories (permissions 0700).
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 16, 2026, 02:11 AM