Run a FlareDeck agent on your own machine: it picks up cards assigned to it, works each one in a checkout of its own, and reports back on the card.
Flagged — allowed with a warning
Allowed by default policy, but 20 findings warrant review before installing.
AI security review
scanned · by lpm-firewall-ai
The package provides a real remote-shell capability that runs manager-supplied commands as the local user and returns their output. Because that capability is enabled by default and requires a separate local opt-out, it presents a dangerous capability users should know about, though the reviewed code routes it through the product’s intended FlareDeck workflow.
Static reason
One or more suspicious static signals were detected.; previous stored version diff introduced dangerous source
Trigger
The agent is running and receives a FlareDeck `exec` event; execution proceeds unless the local remote-off switch is set.
Impact
A manager able to issue an exec request can use the agent’s local user permissions and receive command output. The capability is functional but enabled without a local opt-in.
Mechanism
The agent claims the request using its authenticated FlareDeck account, runs the supplied command with `bash -lc` in the configured repository, then posts the result to the configured FlareDeck API.
The package exposes a CLI that starts the agent from dist/cli.js.
The running agent handles FlareDeck `exec` events using its authenticated account.
A claimed manager command is run through `bash -lc`, and its output is posted back to the configured FlareDeck API.
Remote commands are enabled by default; the user can disable them with `flaredeck-agent remote off`.
Evidence against
Remote execution is described as a workspace manager feature, with requests claimed through the agent’s FlareDeck account.
The command runner bounds execution time and retained output, and a local switch refuses commands when disabled.
The package description identifies it as an agent that works assigned cards and reports back; the remote command path is part of that workspace-agent functionality.
Behavioral surface
SourceChild process · Crypto · Environment variables · Eval · Filesystem · Network · Shell · Web socket
Supply chainHigh entropy strings · URL strings
ManifestNo license
Scanned 133 files · 1.65 MB of source
14 external domains referenced
127.0.0.1
api.github.com
app.flaredeck.app
flaredeck-agent-861259873190.us-central1.run.app
generativelanguage.googleapis.com
github.com
jam.dev
mcp.cnjure.design
mcp.flaredeck.app
mcp.jam.dev
mcp.sceneflo.ai
openrouter.ai
pypi.org
registry.npmjs.org
Findings
1 critical · 6 high · 5 medium · 8 low
Where they are, most severe file first. Code opens a file's flagged lines, with their evidence, below.
382 { type: 'text', text: `This is the screen at ${viewport.width}×${viewport.height}. Judge only what is plain to see; never give a size in pixels.${takeStuckNote()}` },
// npx puts npm[redacted] in the environment, and the environment beats the config FILE
32// we just wrote — which is why `npm config set prefix` appeared to work and the install
⋯
58functionshellProfile() {
59const home = homedir();
60const shell = process.env['SHELL'] ?? '';
61const candidates = shell.includes('zsh')
62 ? ['.zshrc', '.zprofile', '.profile']
63 : ['.bashrc', '.bash_profile', '.profile'];
'never'
35});
⋯
44}
45/** `https://github.com/Owner/Repo(.git)` → `owner__repo`: the clone's folder name. */