AWI

Security checks across static analysis, malware telemetry, and agentic risk

Overview

AWI is a coherent web-reading/search CLI skill, with the main cautions being its user-installed GitHub release binary and agent-executed web requests.

Before installing, make sure you trust the AWI GitHub release because the installer downloads and runs a prebuilt binary. Also be aware that the tool makes web requests, can use stealth/browser-rendering fallback, and supports batch URL reads; use it only for sites and URLs you are comfortable accessing through the agent.

Static analysis

No static analysis findings were reported for this release.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Risk analysis

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.

#ASI04: Agentic Supply Chain Vulnerabilities

Low

What this means

Installing the skill requires trusting the GitHub release asset for the awi binary.

Why it was flagged

The installer fetches a prebuilt executable from GitHub Releases. This is disclosed and purpose-aligned, but no checksum or signature verification is shown.

Skill content

URL="https://github.com/${REPO}/releases/download/${VERSION}/${BINARY}" ... curl -fSL "$URL" -o "${INSTALL_DIR}/awi"

Recommendation

Install only if you trust the repository and release; ideally verify checksums or signatures if the publisher provides them.

#ASI05: Unexpected Code Execution

Low

What this means

The downloaded executable will run on the user's machine during installation verification and later when the skill is used.

Why it was flagged

The installer makes the downloaded binary executable and runs it once to verify installation. This is normal for a CLI installer, but it is still local code execution from an external binary.

Skill content

chmod +x "${INSTALL_DIR}/awi" ... if "${INSTALL_DIR}/awi" --help &>/dev/null; then

Recommendation

Review the source/release provenance before running the installer, and install it under a user account without unnecessary privileges.

#ASI02: Tool Misuse and Exploitation

Info

What this means

The agent can make web requests through the awi command, including reading multiple URLs from a file if directed.

Why it was flagged

The skill allows command execution so the agent can run the awi CLI for URL reading, search, and batch URL processing. This is aligned with the web-interface purpose.

Skill content

allowed-tools:
  - exec
...
awi read "https://example.com/article"
awi search "AI agent frameworks"
awi batch urls.txt

Recommendation

Use the tool for intended web-reading/search tasks, and avoid sending sensitive internal URLs or large batch lists unless you are comfortable with those requests being made.