octocode-pull-request-reviewer

Pass

Audited by Gen Agent Trust Hub on Jun 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes shell git commands (git status, git diff, git branch, git log) to identify changes and retrieve diffs from the local working tree and staged area.
  • [EXTERNAL_DOWNLOADS]: Visual assets and configuration documentation are fetched from the author's public GitHub repository (bgauryy/octocode-mcp), which is a well-known service.
  • [REMOTE_CODE_EXECUTION]: The installation instructions recommend using npx to execute the octocode-mcp server and the npx add-skill command for installation, which is a standard method for Node.js-based agent tools.
  • [PROMPT_INJECTION]: The skill processes untrusted data from external sources such as Pull Request code, developer comments, and local file changes, creating an attack surface for indirect prompt injection.
  • Ingestion points: Data enters the context through githubSearchPullRequests (comments/metadata), git diff, and localGetFileContent (source code).
  • Boundary markers: Instructions do not specify explicit delimiters or "ignore embedded instructions" warnings when interpolating code changes into analysis prompts.
  • Capability inventory: The agent has access to shell execution (git), filesystem write permissions (.octocode/ report directory), and network connectivity via the GitHub API tools.
  • Sanitization: No sanitization or validation of the ingested code content is performed before it is analyzed by the agent.
  • [SAFE]: No obfuscation, hidden URLs, unauthorized data exfiltration, or unauthorized privilege escalation attempts were found in the skill's instructions or logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 18, 2026, 04:44 AM
Security Audit — agent-trust-hub — octocode-pull-request-reviewer