Skillv1.0.0

ClawScan security

system memory inspector · ClawHub's context-aware review of the artifact, metadata, and declared behavior.

Scanner verdict

ReviewMar 13, 2026, 6:56 AM

Verdict: Review
Confidence: medium
Model: gpt-5-mini
Summary: The skill's stated purpose (system-wide memory leak inspection) matches its instructions, but it omits important operational and privilege details and persists potentially sensitive per-process command lines to /var/log without describing required privileges or data handling — exercise caution before installing.
Guidance: This skill implements a system-wide memory scanner that reads /proc for every process and writes snapshots and per-PID trend files under /var/log/memory-inspector; before installing, verify the following: (1) Privileges: to see other users' processes and to write /var/log you likely must run the script as root — decide if you are comfortable granting that. (2) Sensitive data: the script stores truncated process command lines (cmdline), which can contain secrets; remove or sanitize cmdline capture if you cannot guarantee sensitive arguments won't be logged. (3) Persistence and retention: the skill recommends cron scheduling and keeps trend/history files — set log rotation, file ownership, and access controls to prevent unauthorized reading. (4) Installation ambiguity: there is no provided installer; the script is embedded in the documentation, so ensure the script is sourced from a trusted place and review it before creating/executing files. (5) Test in a non-production environment first to confirm behavior and resource impact. If you need help making the script safer, consider restricting it to user-owned processes, writing data to a user-writable directory, or redacting cmdline contents before persisting.

Review Dimensions

Purpose & Capability: noteName and description match the instructions: the SKILL.md defines a scanner that reads /proc, records RSS/VmSize over time, and produces reports. However the skill does not declare that it needs elevated privileges or file-system write access (it writes under /var/log/memory-inspector and suggests installing a cron job), which is a practical requirement for full-system coverage and should have been stated.
Instruction Scope: concernInstructions explicitly read /proc/<pid>/status and /proc/<pid>/cmdline for all PIDs, persist snapshots and per-pid trend files under /var/log, and recommend cron scheduling. Persisting cmdline and process metadata can capture sensitive command-line arguments (passwords, tokens). The SKILL.md does not mention sanitization, access controls, or retention/rotation policies. It also assumes the provided shell script will be created/executed, but there is no install step to place the script on disk — operational ambiguity.
Install Mechanism: noteNo install spec is provided (instruction-only). That lowers supply-chain risk, but also means the user/agent must create the script manually if they follow the instructions. The SKILL.md includes the script content inline, so correctness depends on whoever writes/installs it; there's no signed/release artifact or package referenced.
Credentials: concernThe skill requests no environment variables or external credentials, which is consistent. However it implicitly requires filesystem write access to /var/log and the ability to read other users' /proc entries to achieve full-system inspection. Those privilege needs are not declared. Also collecting cmdline arguments is disproportionate to a minimal memory-only summary and increases risk of capturing secrets.
Persistence & Privilege: concernThe skill suggests persistent presence via cron scheduling and writing persistent snapshot/trending files under /var/log/memory-inspector. Although always:false, the instructions create long-lived files and cron entries which require appropriate privileges; the skill does not document required permissions, log rotation, or access controls for these artifacts.