Claw Claude
PassAudited by ClawScan on May 13, 2026.
Overview
This is a coherent Claude Code integration, but users should understand it delegates coding work to an external CLI that can modify projects, use a logged-in Claude account, and run long tasks or subagents.
Install only if you want coding requests routed through the Claude Code CLI. Before using it on important repositories, make sure Claude Code is installed from a trusted source, confirm which Anthropic account is logged in, use plan mode for larger changes, and monitor long-running or parallel subagent tasks.
Findings (5)
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.
Claude Code may read, modify, build, test, or scaffold local code projects when the assistant uses this skill.
The skill intentionally routes code mutation and review through an external CLI instead of the platform's built-in tools. This is its disclosed purpose, but it means the CLI can act on local project files.
All code-related work goes through Claude Code CLI (`claude`) ... Never use built-in tools (Write / Edit / Bash code-modification commands) to directly manipulate source files.
Use it for projects where you are comfortable giving Claude Code write access, and ask for plan mode or explicit confirmation before large changes.
Long or parallel coding tasks could consume CPU, memory, terminal sessions, or make overlapping file changes if scoped poorly.
The skill documents background/parallel subagent use and no timeout for those invocations. This is disclosed and task-oriented, but it can keep local work running for a long time.
Long-running task | Background subagent, main session stays responsive ... Subagents inherit the same no-timeout policy — they run to natural completion
Limit subagents to clearly independent tasks, monitor long runs, and stop them manually if they exceed expectations.
The skill may use the Claude Code CLI under your logged-in Anthropic account.
The integration expects the user to install Claude Code and authenticate with Anthropic. Account access is normal for this provider integration, and the artifacts do not show credential logging or unrelated use.
npm install -g @anthropic-ai/claude-code ... claude login
Install Claude Code from the official source, understand which account is logged in, and review Anthropic/Claude Code permissions and billing behavior.
A future upstream package or repository change could alter what gets installed if users follow the README manually.
The README documents installation from an external GitHub repository and a global npm install. That is common for this type of CLI skill, but versions are not pinned in the documentation.
git clone https://github.com/dajiaohuang/claude-code-skill.git
Prefer official Claude Code installation instructions, verify the repository before cloning, and pin or review versions for sensitive environments.
Project-level instructions or resumed session context may influence later coding work.
The workflow can create persistent project instructions and resume prior Claude Code sessions. This is normal Claude Code behavior, but persistent context can affect future tasks.
run `claude init` to generate `CLAUDE.md`, ensuring all subsequent changes go through Claude Code ... `--resume` Resume last interactive session
Review generated `CLAUDE.md` files and be careful when resuming sessions in repositories that may contain untrusted instructions.