BargLabs/dunstan

github analyzed

Checks a coding agent's handback against the pull request's GitHub record. No model decides.

maintainer
BargLabs
licence
AGPL-3.0
first seen
2026-10-10
last seen
2026-10-10
releases · 30d
9
github stars
0 reported
short id

Last release 1d ago · 9 releases in 90 days

No dependencies recorded yet.

lifetime releases
9
releases · 30d
9
dependencies
0
dependents
0

what we found

Reading the source raised 2 review prompts — dangerous code and toxic flow (lethal trifecta) — each a pattern worth a human look rather than a finding of fault.

Its 3 tools appear to reach filesystem, shell / exec, network and untrusted content, inferred from source code and tool definitions rather than from observed behaviour.

full security breakdown →

Run it

npx -y dunstan@0.1.22 <arg> reported

{
  "mcpServers": {
    "dunstan": {
      "command": "npx",
      "args": [
        "-y",
        "dunstan@0.1.22",
        "<arg>"
      ],
      "env": {
        "GITHUB_TOKEN": "<your value>",
        "GITLAB_TOKEN": "<your value>"
      }
    }
  }
}
name READMEcode
DUNSTAN_READ_TOKEN no yes
GITHUB_TOKEN yes yes
GITLAB_TOKEN yes no

The list may be incomplete.

The code reads the environment without naming some variables; those cannot be listed.

source: plugins/dunstan/.mcp.json line 4 · scanned 2026-11-16 · analyzer v38

tools

3
  • dunstan_check_handback
  • dunstan_suggest_declarations
  • dunstan_template_handback

release cadence · 90d

9 releases
2026-08-19 2026-10-02 now

recent releases

last 9
version date src
v0.1.22 2026-11-14 github
v0.1.21 2026-11-11 github
v0.1.20 2026-11-11 github
v0.1.19 2026-11-11 github
v0.1.18 2026-11-11 github
v0.1.16 2026-11-03 github
v0.1.15 2026-11-03 github
v0.1.14 2026-11-03 github
v0.1.13 2026-10-23 github

view all →

BargLabs/dunstan is an MCP server distributed on github, maintained by BargLabs, tracked here since October 2026. It has shipped 9 releases and exposes 3 tools. Tools include dunstan_check_handback, dunstan_suggest_declarations, dunstan_template_handback. Its composite risk grade is B — an inferred review prompt computed from observed signals, not a verdict. Setup details are listed below.