Sequential Thinking

Structured reasoning through sequential thinking — break complex problems into steps, solve each independently, verify consistency, synthesize conclusions wi...

Security Scan

The skill's code and runtime instructions are consistent with its stated purpose (structured, stepwise reasoning) and request only the OpenRouter API key which is reasonable for its behavior.

Purpose & Capability

Name/description describe stepwise decomposition and verification; the included Python script implements exactly that flow (decompose → solve steps → optional verify → synthesize) and only requires an OpenRouter API key. No unrelated capabilities or credentials are requested.

Instruction Scope

Runtime instructions and the script send user questions and intermediate step content to OpenRouter's chat completion endpoint — this is expected for an LLM-based reasoning skill but means user data (including intermediate reasoning) is transmitted to a third party. The script also attempts to parse JSON from model responses and falls back to regex extraction, which is normal but can fail if the model output is malformed.

Install Mechanism

No install spec — instruction-only plus a small Python script. The only external dependency is the requests Python package (the script prints a helpful error if requests is missing). There are no downloads from arbitrary URLs or archive extraction.

Credentials

Only OPENROUTER_API_KEY is required and declared as the primary credential, which matches the script's use. No additional secrets, config paths, or unrelated credentials are requested.

Persistence & Privilege

The skill does not request always:true, does not modify other skills or system-wide settings, and does not persist extra privileges; it runs as a one-off script invoked by the agent/user.

Assessment

This skill is coherent with its purpose, but be aware that all questions and intermediate reasoning are sent to OpenRouter (so do not include secrets or private data in prompts). Ensure you trust the OpenRouter service and the holder of OPENROUTER_API_KEY; rotate the key if it is shared. Install the requests Python package in a controlled environment (pip install requests). Review the script if you need to limit which model names are allowed (the --model flag accepts arbitrary model identifiers) to reduce the risk of accidentally using an untrusted model endpoint. Finally, test the skill with non-sensitive inputs first to confirm behavior and outputs.

Structured reasoning through sequential thinking. Break complex problems into logical steps, solve each independently, verify consistency, and synthesize a final answer with a confidence score.

Why Sequential Thinking?

LLMs often rush to conclusions. This skill forces step-by-step decomposition:

1. Decompose — Break the problem into discrete steps
2. Solve — Address each step independently
3. Verify — Check consistency between steps
4. Synthesize — Combine into a final answer with confidence

Usage

# Basic sequential reasoning
python3 {baseDir}/scripts/sequential_think.py "What would happen to Earth's climate if the Moon disappeared?"

# Limit to 5 steps
python3 {baseDir}/scripts/sequential_think.py "Design a sustainable city for 1M people" --steps 5

# Enable self-verification
python3 {baseDir}/scripts/sequential_think.py "Is P=NP?" --verify

# Use a specific model
python3 {baseDir}/scripts/sequential_think.py "Explain quantum computing" --model anthropic/claude-sonnet-4

# JSON output
python3 {baseDir}/scripts/sequential_think.py "Compare React vs Vue" --json

# Verbose mode (show all intermediate reasoning)
python3 {baseDir}/scripts/sequential_think.py "Solve this logic puzzle..." --verbose

Flags

Flag	Default	Description
--steps	7	Maximum number of reasoning steps
--verify	off	Enable self-verification pass
--model	anthropic/claude-sonnet-4	Model to use
--json	off	Output structured JSON
--verbose	off	Show full intermediate reasoning
--temperature	0.3	Temperature for reasoning (lower = more focused)

Output Format

🧩 Sequential Thinking: "Your question here"
══════════════════════════════════════════

Step 1/5: [Step Title]
  → [Reasoning and conclusion for this step]

Step 2/5: [Step Title]
  → [Reasoning and conclusion for this step]

...

✅ Verification: [Pass/Fail — consistency notes]

📋 Synthesis:
  [Final combined answer]

🎯 Confidence: 85% (High)

How It Works

1. Decomposition prompt asks the model to identify the key sub-questions
2. Step-solving prompts address each sub-question with context from prior steps
3. Verification prompt (optional) checks for contradictions between steps
4. Synthesis prompt combines all step conclusions into a coherent answer
5. Confidence scoring based on step agreement, verification results, and hedging language

Credits

Built by M. Abidi | agxntsix.ai YouTube | GitHub Part of the AgxntSix Skill Suite for OpenClaw agents.

📅 Need help setting up OpenClaw for your business? Book a free consultation