We recommend always running the latest published version:

npx @supermldev/smart-sdlc@latest init

Please do not open a public GitHub Issue for security vulnerabilities.

Report security issues privately by emailing:

security@superml.org

Include in your report:

• A description of the vulnerability and its potential impact
• Steps to reproduce (or a proof of concept)
• Smart SDLC version affected (npx @supermldev/smart-sdlc --version)
• Any suggested mitigations if you have them

We will acknowledge your report within 48 hours and aim to release a fix within 14 days for confirmed critical issues.

You are welcome to request credit in the release notes. We will not take legal action against good-faith security researchers.

• _superml/config.yml is gitignored by default — it stores API tokens, base URLs, and auth headers for integrations
• _superml/persona.yml is gitignored by default — it stores personal preferences, never credentials
• .vscode/mcp.json may contain MCP server tokens — Smart SDLC skills instruct users to add it to .gitignore; it is not written to git-tracked files by any Smart SDLC command

Smart SDLC skills are markdown instruction files read by the AI assistant. The CLI (bin/superml.js) does not execute AI-generated code or evaluate untrusted input.

The CLI itself makes no outbound network requests. All integration calls (JIRA, Confluence, etc.) are made by the AI assistant via curl in a terminal or via MCP tool calls — under the user's own credentials and with explicit user intent.

Skills are static markdown files. If you modify a skill file or add content to _superml/reference/ from an untrusted source, be aware that an attacker-controlled file loaded into AI context could attempt to manipulate the AI's behaviour. Only load reference docs and skills from sources you trust.

Smart SDLC has zero runtime npm dependencies. The attack surface from the installed package is limited to the CLI source code itself.

The following are in scope for security reports:

• Vulnerabilities in bin/superml.js, lib/, or config/ that could lead to credential exposure, arbitrary code execution, or privilege escalation
• Skill instructions that could be exploited to exfiltrate secrets from the AI's context
• .gitignore generation bugs that could cause credentials to be committed

The following are out of scope:

• Vulnerabilities in the AI assistant itself (GitHub Copilot, Claude, etc.) — report those to the respective vendor
• Vulnerabilities in third-party MCP servers (@sooperset/mcp-atlassian, Atlassian Remote MCP, etc.) — report those to their maintainers
• Social engineering attacks that require physical access to the user's machine

Join the Smart SDLC community on Discord: discord.gg/26XMXkEt

For general questions and discussions, please use Discord rather than GitHub Issues. Security reports should still go to security@superml.org.