← Back to Skills Marketplace
360
Downloads
0
Stars
0
Active Installs
3
Versions
Install in OpenClaw
/install ask-claude-skill
Description
Delegate a task to Claude Code CLI and immediately report the result back in chat. Supports persistent sessions — Claude Code remembers previous context with...
README (SKILL.md)
Ask Claude — Execute & Report (with persistent sessions)
The Two Modes
New session (default)
Use when starting a fresh task or new topic.
OUTPUT=$(/home/xmanel/.openclaw/workspace/run-claude.sh "prompt" "/workdir")
echo "$OUTPUT"
Continue session (--continue)
Use when the user is following up on a previous Claude task in the same workdir. Claude Code will have full memory of what was done before — files read, edits made, context gathered.
OUTPUT=$(/home/xmanel/.openclaw/workspace/run-claude.sh --continue "prompt" "/workdir")
echo "$OUTPUT"
When to use --continue
Use --continue when the user says things like:
- "agora corrige o que encontraste"
- "continua"
- "e o ficheiro X?"
- "faz o mesmo para..."
- "e agora?"
- "ok, e o erro de..."
- Anything that clearly references what Claude just did
Use a new session when:
- New unrelated task
- User says "começa do zero" / "new task" / "esquece o anterior"
- Different workdir/project
Session storage
Claude Code stores sessions per-directory in ~/.claude/projects/.
As long as you use the same workdir, --continue picks up exactly where it left off —
same file context, same conversation history, same edits.
Direct command (alternative to wrapper)
# New session
OUTPUT=$(cd /workdir && env -u CLAUDECODE claude --permission-mode bypassPermissions --print "task" 2>&1)
# Continue session
OUTPUT=$(cd /workdir && env -u CLAUDECODE claude --permission-mode bypassPermissions --print --continue "task" 2>&1)
Common workdirs
| Context | Workdir |
|---|---|
| General/scripts | /home/xmanel/.openclaw/workspace |
| Trading | /home/xmanel/.openclaw/workspace/hyperliquid |
After receiving output
- Summarize in 1-3 lines what Claude did/found
- Mention files created or edited
- If error: analyze and suggest fix
- If output is long: summarize, offer full output on request
Usage Guidance
This skill is coherent with its stated purpose, but review these before installing: 1) Confirm you want the agent to run your local `claude` CLI — anything in the chosen workdir can be read by Claude and returned in chat. 2) Check and, if needed, change the default workdir (/home/xmanel/.openclaw/workspace) used by the wrapper so it doesn't point at a directory containing secrets. 3) Be aware the wrapper passes `--permission-mode bypassPermissions` to the CLI — verify what that flag does for your installed claude CLI/version. 4) Install and trust the official Claude Code CLI from its vendor; verify authentication (claude login) is performed separately. If you want to limit risk, run the skill in an isolated workdir with no sensitive files or avoid using --continue/persistent sessions.
Capability Analysis
Type: OpenClaw Skill
Name: ask-claude-skill
Version: 1.1.1
The skill automates the Claude Code CLI using the `--permission-mode bypassPermissions` flag in `run-claude.sh` and `SKILL.md`, which allows the sub-agent to execute arbitrary shell commands and file modifications without user confirmation. It contains hardcoded absolute paths for a specific user (`/home/xmanel`) and explicitly suggests a workdir for 'Trading' (`hyperliquid`), which is a high-value target. While there is no evidence of intentional exfiltration or malware, the bypass of security prompts combined with access to sensitive directories poses a high risk of unintended autonomous actions or exploitation via prompt injection.
Capability Assessment
Purpose & Capability
Name/description match the implementation: the SKILL.md and run-claude.sh invoke the local `claude` CLI, support --continue for persistent sessions, and capture stdout/stderr. Requiring the claude binary is proportionate to the stated purpose.
Instruction Scope
Instructions explicitly tell the agent to run the claude CLI (or the run-claude.sh wrapper), unset CLAUDECODE, and use `--permission-mode bypassPermissions`. They also document session storage under ~/.claude/projects/ and default workdirs under /home/xmanel. This is coherent with the skill's purpose, but means Claude (via the CLI) will have access to files in the chosen workdir and any session history stored in the user's home — which can expose secrets from those directories if present.
Install Mechanism
No install spec is present (instruction-only plus a small wrapper script). Nothing is downloaded or written by an installer; the only runtime requirement is that the official `claude` CLI is installed and authenticated.
Credentials
The skill declares no required env vars (appropriate). The instructions and wrapper intentionally unset CLAUDECODE (env -u CLAUDECODE) before launching claude; this is consistent with avoiding nested-session errors but is worth noting. No unexpected credentials or unrelated environment access are requested.
Persistence & Privilege
The skill uses Claude Code's persistent sessions (per-directory storage in ~/.claude/projects/) to implement --continue behavior. The skill itself does not request always:true and does not modify other skills. However, persistent sessions mean the agent (via Claude) can access historical session data and files within the workdir, increasing the scope of what is read and potentially reported back.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install ask-claude-skill - After installation, invoke the skill by name or use
/ask-claude-skill - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.1.1
Updated README with full documentation: feature table, persistent sessions explained, v1.0.0 vs v1.1.0 changelog, usage examples with continue/new session decision logic.
v1.1.0
- Adds persistent session support: Claude Code now remembers previous context within the same workdir, enabling follow-ups using --continue.
- New run-claude.sh wrapper script added for handling session management.
- Updated usage instructions and trigger phrases to explain when to start a new session or continue a previous one.
- Output handling guidance improved: summarize results, mention files affected, and advise on errors.
- Documentation reorganized for clarity on modes, session storage, and common workdirs.
v1.0.0
- Initial release of the ask-claude skill.
- Allows users to delegate code execution and analysis tasks to the local Claude Code CLI directly from OpenClaw.
- Always runs tasks synchronously and returns the full result in chat, following a strict output pattern.
- error handling, file edits, and long outputs are explicitly addressed in the usage rules.
- Skill auto-triggers on various task delegation phrases like "ask claude to..." or "use claude code to...".
Metadata
Frequently Asked Questions
What is Ask Claude Skill?
Delegate a task to Claude Code CLI and immediately report the result back in chat. Supports persistent sessions — Claude Code remembers previous context with... It is an AI Agent Skill for Claude Code / OpenClaw, with 360 downloads so far.
How do I install Ask Claude Skill?
Run "/install ask-claude-skill" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Ask Claude Skill free?
Yes, Ask Claude Skill is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Ask Claude Skill support?
Ask Claude Skill is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Ask Claude Skill?
It is built and maintained by 0xManel (@0xmanel); the current version is v1.1.1.
More Skills