xkcd
Pass
Audited by Gen Agent Trust Hub on Nov 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill's setup process in
runtime/scripts/dependencies.mjsperforms an automated installation of Node.js dependencies from the npm registry usingnpm ci. It also downloads the Chromium browser via Playwright when browser-based features are first used. - [COMMAND_EXECUTION]: The skill relies on system binaries to perform its core functions, executing
gitto retrieve source code,gh(GitHub CLI) to interact with pull requests, andnodeto run its internal management scripts. - [DATA_EXFILTRATION]: The skill transmits technical summaries and generated comic content to GitHub as PR comments using the
gh api. To mitigate the risk of leaking sensitive information, theredactutility inruntime/scripts/core.mjsidentifies and masks patterns matching API keys, tokens, and private keys. It also explicitly excludes.envfiles from code capture. - [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes source code from potentially untrusted external repositories. While it instructs the agent to treat code as data and includes a redaction layer, it remains exposed to indirect prompt injection if malicious instructions are hidden within the analyzed code.
- Ingestion points: Source code is ingested through
captureLocalandcapturePrinruntime/scripts/core.mjs. - Boundary markers:
runtime/guide.mdexplicitly directs the agent to treat code as data, not instructions. - Capability inventory: The skill can execute shell commands (
git,gh,npm), perform file system operations, and interact with the GitHub API. - Sanitization: The
redactfunction inruntime/scripts/core.mjsprovides automated filtering of hardcoded credentials before they are processed or exported.
Audit Metadata