superdocs

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The prompt explicitly instructs the agent to include "config values" and "code examples from the actual codebase" (actual paths, commands, config values), which may force the LLM to copy secret values verbatim from repository/config files into generated outputs, creating an exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The repo deliberately invokes an external "claude" CLI to run an automated agent with broad Read/Write/Edit and shell tool access over the target project (and explicitly forbids interactive confirmation), which creates a high risk of intentional data exfiltration and remote execution by an untrusted remote model.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). High likelihood: the required workflow reads and ingests existing repository files (e.g., README.md, CLAUDE.md/AGENTS.md, CONTRIBUTING.md, CHANGELOG.md, scattered ADRs, and docs/ in MIGRATE/UPDATE) into the LLM context via the generated prompt and tool reads, and those files may be outsider-authored (e.g., from a PR/issue branch or third-party repo content).