Skillv1.1.0

ClawScan security

Competitor Watch · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

BenignFeb 11, 2026, 8:42 PM

Verdict: benign
Confidence: medium
Model: gpt-5-mini
Summary: The skill's code and instructions are consistent with a website monitoring/competitive-intelligence tool, but there are a few missing declarations and small mismatches you should review before installing.
Guidance: This skill appears to do what it says, but review these before installing: 1) Dependencies: ensure jq and common Unix tools (diff, grep, sed, awk, uuidgen, base64) are available — the scripts will fail otherwise. 2) Alerts: decide which notification channel you'll use (Telegram, email, Slack, etc.) and configure/store any tokens/webhooks securely; the example config uses 'telegram' but no token handling is provided. 3) Data storage & privacy: snapshots store full page content — consider retention settings and legal/privacy implications of scraping competitor pages (and whether snapshots could contain customer PII). 4) Dry-run & audit: run scripts in --dry-run first, inspect the created files under ~/.config/competitor-watch, and review report output. 5) Execution scope: the scripts assume a 'clawd' fetch/message tool exists or will be implemented — replace or implement those calls safely. If you need higher assurance, ask the author to: declare required binaries and any credential env vars, document alert-channel credential setup, and remove or explain any use of random/base64/uuid utilities. If anything looks unexpected after those changes, re-evaluate.

Review Dimensions

Purpose & Capability: okName/description match the files and behavior: scripts implement adding competitors, fetching pages, taking snapshots, diffing, scoring, and generating alerts. Required config paths (~/.config/competitor-watch) and data storage are coherent with the stated purpose.
Instruction Scope: noteThe SKILL.md and scripts operate entirely within the monitoring domain (fetch pages, store snapshots, compute diffs, log changes and send alerts). They create and write to ~/.config/competitor-watch and its data subdirs (snapshots, change log). This is expected, but be aware snapshots will contain full scraped page contents (possible sensitive or copyrighted content). The scripts include a simulated 'clawd' fetch/send command — in a real deployment you must provide the actual fetch/notification tooling.
Install Mechanism: okNo remote install/downloads are performed by the skill (instruction-only with local scripts). There are no URLs or archive extracts. Scripts are local and executed by the user — lower install risk than remote downloads.
Credentials: noterequires.env is empty and no credentials are listed, which aligns with a simple web-scraping approach; however: (1) the scripts require external CLI tools (jq is checked at runtime, and utilities like diff, grep, sed, awk, uuidgen, base64 are used) but these are not declared in metadata; (2) alerts.channel defaults to 'telegram' in the example config but no guidance or env var for Telegram bot tokens/webhook is provided — you'll need to supply and secure those credentials if you enable that channel.
Persistence & Privilege: okThe skill does not request 'always: true' or other elevated platform privileges. It writes only to its own config/data under the user's home directory (or CW_CONFIG_DIR if set). It does not modify other skills or system-wide settings.