Skip to content

Security: aithinkers/devloop

Security

SECURITY.md

Security

DevLoop is a local, dependency-light toolkit (bash + python3 + git). It runs entirely on your machine and makes no network calls of its own — model calls are made by your host (Claude Code / Kiro / Codex), and DevLoop never writes to Jira or any external service (the Jira Organizer produces guidance + config only). Still, a few things are worth knowing.

Install from the official source only

The bootstrap is a curl … | sh that clones and runs code, so only install from the official repository — https://github.com/aithinkers/devloop (or a clone you've reviewed). Don't pipe an installer from a fork or mirror you don't trust. You can always read install.sh and devloop first; both are short, dependency-free shell.

Threat model & guidance

  • Ingesting documents (ingest.py). It parses office/zip/XML files (.docx/.pptx/.xlsx, .drawio, .vsdx, .svg) with the Python standard library. Parsing untrusted files is an attack surface (e.g. a hostile XML could attempt resource exhaustion). Run ingest.py on sources you trust; treat third-party documents with the same caution you'd give any untrusted input. It copies/extracts text locally and never uploads anything.
  • Git-backed wikis (wikikit.py sync). This clones the git URLs you put in devloop.wikis.json. Only add repositories you trust.
  • Helper execution. wikikit.py and ingest.py are pure-stdlib scripts that run locally under your host's normal permission model; review them if your environment requires it.
  • Secrets. DevLoop ships no credentials. The MCP examples reference secrets via environment variables (e.g. ${SHAREPOINT_TOKEN}) — never hardcode tokens into configs you commit.

Reporting a vulnerability

Please report security issues privately via a GitHub security advisory rather than a public issue, and allow time for a fix before disclosure.

There aren't any published security advisories