X Reply Unreplied

Check X notifications, mentions, or reply threads and answer posts that still need a response from the user. Use when the user wants help replying to unrepli...

Security Scan

The skill's instructions and requirements are broadly consistent with its stated purpose (helping a user reply to X notifications), but it assumes an authenticated OpenClaw-managed browser session and runtime permissions to read and post on the account — make sure you understand and approve that operational context before use.

Purpose & Capability

The name and description match the instructions: the skill is explicitly about reviewing X (formerly Twitter) notifications and posting short replies. It reasonably requires browser interaction and an authenticated account, but the registry metadata lists no required binaries or credentials even though the SKILL.md mandates using the 'openclaw browser start' command and posting as the user. This is a minor mismatch: the skill assumes the platform provides the managed browser and an active logged-in session but does not declare that dependency.

Instruction Scope

The SKILL.md stays on-topic: it instructs the agent to open the managed browser, navigate X notifications, read threads, draft one- or two-sentence replies, and post them. It does not ask to read unrelated files or environment variables. It does require the agent to like posts and to post replies on the user's behalf, and enforces non-refresh and sequence rules — these are operational constraints rather than scope creep.

Install Mechanism

This is an instruction-only skill with no install spec or code to download — low install risk. Nothing will be written to disk by an installer as part of this skill.

Credentials

The skill declares no environment variables or credentials, which is coherent for an instruction-only skill. However, it implicitly requires an authenticated browser session (cookies/credentials stored in the managed browser) to post on the user's behalf. The skill does not request or document how authentication is provided; users should verify the managed browser will be logged into the correct account and that they are comfortable allowing the agent to act in that session.

Persistence & Privilege

The skill is not marked always:true and does not request persistent system modifications or cross-skill configuration changes. It is user-invocable and the agent can invoke it autonomously (platform default), which is expected for this type of automation.

Assessment

This skill appears to do what it says: review X notifications and post very short replies. Before installing, confirm the platform provides the OpenClaw-managed browser and that it will be logged into the correct X account — the skill will act in that browser session and post as you. If you do not want automated posting, only use the skill manually or review drafts before posting. Also verify you are comfortable with the 'like before drafting' and non-refresh rules. If you need more assurance, ask the developer to declare the browser dependency explicitly and to document how authentication is handled.

Use this skill for manual reply handling, not a scheduled inbox sweep. It assumes browser access and works best when a twitter-humanizer skill is also available.

Do not use this skill to evade platform enforcement, simulate human behavior for detection avoidance, or mass-reply at scale.

If the user did not specify a scope, default to notifications and mentions that appear to expect a reply. If the user did not specify a count limit, process up to 5 suitable reply opportunities in one run.

Reply standard

• One thread at a time
• Reply only where the account has not already replied in the same thread
• Stay on the thread topic only
• Use first person
• CRITICAL NON-NEGOTIABLE: keep every reply to 1 or 2 short sentences.
• CRITICAL NON-NEGOTIABLE: never avoid, relax, or reinterpret this limit unless the user explicitly asks for longer replies.
• Do not force product mentions or links unless the user asked for them

Workflow

1. Run openclaw browser start to open the openclaw managed browser. CRITICAL NON-NEGOTIABLE: always use this command. Never open a browser any other way. Then navigate to X and go to notifications or the requested inbox surface.
2. CRITICAL NON-NEGOTIABLE: once on the notifications surface, never refresh or reload the page for any reason for the entire run. Navigate between threads and back using clicks and the back button only.
3. Review candidates one by one.
   • Ignore obvious spam, rage bait, bot replies, or posts that do not need a response.
   • Open the thread and read enough context to understand what the person is reacting to.
   • Confirm the user account has not already replied in-thread.
   • Read references/inbox-workflow.md.
4. Like the post before drafting or touching the reply box.
5. Draft the reply.
   • If the twitter-humanizer skill is available, use it to produce 2 short reply options and pick the better one.
   • Keep the reply friendly, direct, and thread-specific.
   • Read references/reply-rules.md.
6. Post the reply.
7. Return to notifications using the back button and continue until the cap is reached or there are no more good candidates.
8. CRITICAL NON-NEGOTIABLE: close every x.com tab immediately after the run is complete. Do not leave any tab open.

CRITICAL NON-NEGOTIABLE: do not jump across multiple threads in parallel. CRITICAL NON-NEGOTIABLE: finish one full thread cycle (open -> read -> decide -> draft -> post or skip -> return) before opening the next thread.

Skip conditions

Do not reply when:

• The user already answered
• The message is abusive with no upside to engaging
• The post is pure spam or obvious automation
• The right answer would require facts you cannot verify quickly
• A reply would likely derail into support, legal, or security issues better handled elsewhere

Completion report

At the end, report:

• How many threads were reviewed
• Which ones were replied to
• The final text used for each reply
• Which items were skipped and why
• Confirmation that all x.com tabs were closed