Claw Smart Context

PassAudited by ClawScan on May 1, 2026.

Overview

This is an instruction-only efficiency skill with no code or credentials, though it does encourage faster tool use and optional delegation that users may want to understand.

This skill appears benign and purpose-aligned for reducing token and tool usage. Before installing, consider whether you are comfortable with an agent prioritizing efficiency, using command-line-style searches when available, skipping extra checks on low-risk tasks, and delegating some work to trusted sub-agents.

Findings (2)

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.

Low

#ASI02: Tool Misuse and Exploitation

What this means

The agent may use existing command tools more readily and may not double-check simple changes unless the user asks.

Why it was flagged

The skill advises the agent to prefer command execution for some searches and to avoid extra verification for low-risk operations. This is purpose-aligned with efficiency, but users should know it may favor speed and token savings over conservative checking.

Skill content

**Prefer exec over multiple reads** — `grep` across files is cheaper than reading 5 files separately
- **Skip verification for low-risk ops**

Recommendation

For important file changes or sensitive work, explicitly ask the agent to verify results and avoid shell commands unless needed.

Low

#ASI07: Insecure Inter-Agent Communication

What this means

If the host environment supports sub-agents, some work or files may be handed to another agent/session to save cost.

Why it was flagged

The skill encourages delegation to other agents or background sessions when available. This is disclosed and limited by guidance not to delegate context-sensitive tasks, but it can still affect where task data is processed.

Skill content

If sub-agents or background sessions are available, use them with cheaper models for:

- Background research that doesn't need conversation context
- File processing, data formatting, bulk operations

Recommendation

Use this skill in environments where delegated agents are trusted, and tell the agent not to delegate if the task involves sensitive data.