registry / @cotal-ai/cli / 0.135.12
@cotal-ai/cli@0.135.12
Cotal mesh CLI: up, join, watch, console, spawn, mint, channels, history, ext.
AI Security Review
scanned 2d ago · by lpm-firewall-aiThe flagged behaviors are user-invoked CLI features: setup writes the package's documented skill files, while feedback sends a submitted report to Cotal's endpoints. The manifest has no install lifecycle hook, and the inspected code shows no automatic credential exfiltration or unrelated agent-control change.
Static reason
One or more suspicious static signals were detected.
Trigger
Setup, feedback, join, and persona editing run through explicit CLI commands; dispatcher effects require a validated invocation.
Impact
No unrequested credential transfer, unrelated collection, or unauthorized agent-setting change was identified.
Mechanism
Setup reconciles Cotal-authored skills, feedback sends a user-submitted report to a Cotal endpoint by default, and editor/process execution supports CLI workflows.
AI rationale
The flagged behaviors are user-invoked CLI features: setup writes the package's documented skill files, while feedback sends a submitted report to Cotal's endpoints. The manifest has no install lifecycle hook, and the inspected code shows no automatic credential exfiltration or unrelated agent-control change.
Evidence
~/.agents/skills/<name>/SKILL.md~/.cotal
Network endpoints2
broker.cotal.aicotal.ai
Decision evidence
public snapshotAI called this Clean at 95.0% confidence as Benign with low false-positive risk.
Evidence for AI clean decision
- The manifest points to JavaScript entry points and declares build and smoke-test scripts; it has no install lifecycle hook.
- The CLI writes cross-vendor skills only through the explicit setup command, including `setup --skills`.
- Skill reconciliation manages the shipped `SKILL.md` files under `~/.agents/skills`; it checks for symlinks and preserves divergent copies.
- Feedback submission is an explicit command that sends the entered report to the Cotal feedback endpoint; any configured feedback key is attached to that request.
- Dispatcher effects follow a validated command invocation.
Evidence against
- Setup can install the CLI globally when run through npx; the function checks whether the CLI is already available and prompts in interactive use.
- The package writes behavior-bearing skill files into the user's agent skills directory during setup.
- Feedback accepts a user-configured endpoint, so its destination may differ from the two Cotal defaults.
Behavioral surface
ChildProcessCryptoDynamicRequire
Source & flagged code
6 flaggeddist/seed/reconcile.jsView file
1import { existsSync, renameSync } from "node:fs";
L2: import { spawnSync } from "node:child_process";
L3: import { randomBytes } from "node:crypto";
High
Child Process
Package source references child process execution.
dist/seed/reconcile.jsView on unpkg · L1dist/lib/persona-files.jsView file
274process.on(signal, pass);
L275: const child = program ? spawn(program, [draft], { stdio: "inherit" }) : spawn(`${editor} ${quoted}`, { shell: true, stdio: "inherit" });
L276: return new Promise(resolve => {
High
dist/lib/nats-bin.jsView file
Findings
5 High4 Medium4 Low
HighChild Processdist/seed/reconcile.js
HighShelldist/lib/persona-files.js
HighTrigger Reachable External Ai Agent Control Surface Mutationdist/lib/agent-skills.js
HighRuntime Package Installdist/commands/setup.js
HighSemantic Analysis Limiteddist/commands/ext.js
MediumDynamic Requiredist/lib/nats-bin.js
MediumNetwork
MediumEnvironment Vars
MediumStructural Risk Force Deep Review
LowScripts Present
LowFilesystem
LowHigh Entropy Strings
LowUrl Strings
Affected versions and remediation
This report applies to @cotal-ai/cli@0.135.12.
See all versions of @cotal-ai/cli for other recorded verdicts.
- Review the evidence and your use of @cotal-ai/cli@0.135.12 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.