OpenClaw Problem Solver (v3.0)

This skill acts as a dedicated diagnostic, resolution, and validation engine for any question or bug report related to the OpenClaw framework itself. It moves beyond simple search by actively verifying solutions through execution when necessary.

When to Use This Skill

Use this skill when the user:

• Asks "Why is [feature] not working in OpenClaw?"
• Reports a specific bug (e.g., "The gateway tool fails with error X").
• Needs guidance on how to implement a specific feature using OpenClaw's architecture or tools.
• Wants to know the best practice for a certain task within the OpenClaw ecosystem, but requires verification.

Core Workflow: The 6-Step Resolution Cycle (v3.0)

When triggered, this skill must execute the following steps sequentially:

Step 0: Pre-Check & Context Gathering - NEW!

Action: Before any search, check context to tailor the approach and set safety parameters. Checks:

1. Session State: Read ~/proactivity/session-state.md for the last explicit goal or active blocking decision.
2. User Preference: Check USER.md and IDENTITY.md (e.g., preferred documentation source, common project context).
3. Initial Triage & Safety Scan: Determine if the query is about a known task/bug. 关键安全检查： 扫描用户提问，判断是否包含敏感信息（API Key, Secret Token等）。如果包含，则标记为 [RISK: HIGH]。

Step 1: Primary Search - Official Documentation (docs.openclaw.ai)

Action: Use tavily_search with a query focused on the official documentation. Query Focus: The user's exact problem description, potentially refined by Pre-Check context. Parameters:

• query: [User's Problem Description]
• include_answer: true (To get an immediate summary)
• search_depth: "advanced"

Goal: Find a direct answer or a link to the relevant documentation page.

Step 2: Fallback Search - GitHub Issues (github.com/openclaw/openclaw/issues)

Action: If Step 1 yields no satisfactory result, use tavily_search again, targeting GitHub issues. Query Focus: The user's exact problem description, often prefixed with "OpenClaw issue: [Problem]". Parameters:

• query: "[User's Problem Description] OpenClaw issue"
• include_answer: true (To get a summary of the best matching issue)
• search_depth: "advanced"

Goal: Find an existing, reported bug or discussion thread that mirrors the user's problem.

Step 3: Synthesis & Decision (The Core Logic)

Once results are gathered from Step 1 and/or Step 2, perform critical thinking to select the best path forward. Decision Criteria:

• Definitive Answer: Docs provided a clear solution $\rightarrow$ Proceed to Validation (Step 4A).
• Workaround Found: GitHub Issue provides a quick fix $\rightarrow$ Propose workaround and suggest official fix $\rightarrow$ Proceed to Validation (Step 4B).
• Ambiguous/Missing: Both are weak or contradictory $\rightarrow$ Formulate an inquiry based on context $\rightarrow$ Proceed to Contextual Inquiry (Step 5C)。

Step 4: Validation & Action (The Execution Layer) - NEW!

This step executes the chosen path from Step 3. A) Direct Answer: If synthesis is clear, present the solution immediately and conclude. B) Code/Config Verification (MRE): If the problem relates to code or config, proactively call exec with a minimal test case derived from the search results. The output of this execution becomes the primary evidence for the final answer。 C) Contextual Inquiry: If more data is needed, formulate a precise question for the user (e.g., "请问您是在哪个项目目录下运行的？" 或 "能否提供一下报错时的完整日志文件？")。

Step 5: Finalization & Memory Update (The Wrap-up) - NEW!

After the solution is presented or an inquiry is made, this step ensures continuity. Action: Before concluding, update memory based on the outcome:

1. Remember Fact: Store the core problem/solution pair (mem.remember(...)).
2. Learn Lesson: If a complex workaround was needed, log it as a lesson learned (mem.learn(...)).
3. Update State: Update ~/proactivity/session-state.md with the resolution status and next required action。

How to Help Users Solve Problems (Flow Summary)

1. User asks a question.
2. $\rightarrow$ Execute Step 0 (Pre-Check).
3. $\rightarrow$ Execute Step 1 $\rightarrow$ Step 2 (Search Chain).
4. $\rightarrow$ Execute Step 3 (Synthesis & Decision).
5. $\rightarrow$ Execute Step 4 (Validation/Action) based on Step 3's decision.
6. $\rightarrow$ Execute Step 5 (Finalization & Memory Update).

Example Interaction Flow (Enhanced)

User: "我发现用 exec 命令跑一个需要 TTY 的 CLI 时，它总是卡在等待输入界面。"

Agent Action Sequence:

1. Step 0 (Pre-Check): Checks state $\rightarrow$ Sees no active goal, but notes the user is focused on 'Tooling/Exec'.
2. Step 1 Search (Docs): tavily_search(...) $\rightarrow$ Returns summary: "Use pty=true."
3. Step 2 Search (GitHub Issues): tavily_search(...) $\rightarrow$ Returns summary: "Issue #456 confirms this is a known behavior, suggests increasing yieldMs."
4. Step 3 Synthesis: Decision: Docs provide the ideal fix (pty=true), but GitHub provides an immediate workaround (yieldMs). The best path is to propose both and validate with code. $\rightarrow$ Proceed to Validation (Step 4B).
5. Step 4 Validation (MRE): Agent calls exec(command=\"echo 'Waiting for input...' && read -p 'Press Enter to continue:'\", pty=true, yieldMs=100) $\rightarrow$ Exec returns output showing the prompt.
6. Step 5 Finalization: Action: Present solution (Use pty: true first). Memory Update: Remember Fact ("Exec TTY issue solved by pty/yieldMs"). Learn Lesson ("When exec hangs, always check for pty=true or increase yieldMs."). Update State to "Resolved - Awaiting confirmation.".

Final Output: Deliver the synthesized advice clearly!