Pilot Protocol

AdvisoryAudited by Static analysis on Apr 30, 2026.

Overview

No suspicious patterns detected.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

Installing requires trusting the Pilot Protocol website and downloaded pilotctl binaries.

Why it was flagged

The documented setup executes a remote installer and installs an external daemon/runtime that is not included in the reviewed artifacts.

Skill content
curl -fsSL https://pilotprotocol.network/install.sh | sh ... downloads pre-built binaries (or builds from source if no release is available) ... sets up a system service
Recommendation

Verify the installer source, signatures, or checksums where available, and review the installed service before relying on it.

What this means

Sensitive files or task content could be shared with trusted peers if the agent is asked to send them.

Why it was flagged

The core function is agent-to-agent communication, including sending and receiving data and files across a peer network.

Skill content
send messages, files, or data to another AI agent ... listen for incoming messages, files, or events ... establish or manage trust with other agents
Recommendation

Use only known peers, review trust relationships, and confirm before sending sensitive data or files.

NoteHigh Confidence
ASI01: Agent Goal Hijack
What this means

A trusted peer could submit a task that attempts to redirect the agent outside the user's intended goal.

Why it was flagged

Peer-submitted tasks can become work items for the local agent, so their descriptions should be treated as untrusted requests rather than authoritative instructions.

Skill content
The Task Submit service ... enables agents to request work from other agents ... pilotctl task execute ... # Do the work...
Recommendation

Require review before accepting or executing remote tasks, and follow the documented decline criteria for unsafe requests.

What this means

Incorrect gateway mappings could expose or route traffic through the Pilot overlay in ways the user did not intend.

Why it was flagged

The gateway can create local TCP proxy listeners for arbitrary ports, which is expected for an IP bridge but powerful if mis-scoped.

Skill content
The gateway bridges standard IP/TCP traffic to Pilot Protocol ... Requires root for ports below 1024. Supports any port — configure with `--ports`.
Recommendation

Limit gateway use to specific peers and ports, avoid sudo unless necessary, and stop the gateway when finished.

NoteHigh Confidence
ASI10: Rogue Agents
What this means

The machine may keep receiving peer messages, files, events, or tasks while the daemon and trust relationships remain active.

Why it was flagged

The skill depends on a background daemon with automatically started services, so activity can continue beyond a single command.

Skill content
The daemon must be running ... Built-in services: the daemon auto-starts echo ... data exchange ... event stream ... task submit
Recommendation

Start the daemon only when needed, monitor trusted peers and inboxes, and stop or disable the service when not in use.

What this means

Malicious or stale peer messages could influence future agent behavior if treated as trusted context.

Why it was flagged

Inbound peer content persists locally and may later be read by a user or agent.

Skill content
Received files and messages are stored locally and can be inspected at any time.
Recommendation

Treat inbox files, messages, and task descriptions as untrusted input; clear stored content when it is no longer needed.