decision-log

What This Skill Does

The decision-log skill serves as a structured memory bank for the OpenClaw agent, enabling it to document high-stakes choices, the reasoning behind them, and the alternatives that were rejected. By maintaining a rigorous documentation standard, the agent transforms decision-making from a transient event into a trackable asset. The skill features an automated 30-day review cycle that forces a reflection process, comparing anticipated outcomes against actual real-world results. This enables the agent to evolve, learn from historical mistakes, and refine its internal logic over time.

Installation

To integrate this skill into your environment, use the OpenClaw package manager within your terminal: clawhub install openclaw/skills/skills/mupengi-bot/decision-log Ensure your agent has read/write permissions for the memory/decisions/ directory, as the skill generates markdown-based log files for every entry.

Use Cases

This skill is perfect for strategic planning where consistency is key. Use it when managing infrastructure changes, setting marketing guidelines, or adjusting code architecture. For example, if you are deciding to switch database providers, logging the rationale and expected performance metrics allows the agent to evaluate the validity of that decision exactly one month later. It is also highly useful for project management, where tracking why a specific feature was scoped out prevents "re-litigating" the same debate in future sprints.

Example Prompts

1. "Record decision: Migrating the backend to PostgreSQL because our current SQLite setup is hitting write-concurrency limits."
2. "Decision log: Why did I choose to delay the API release? Please document the alternatives and the expected user feedback loop."
3. "Made this decision to prioritize technical debt over new features for the next two weeks to ensure long-term stability."

Tips & Limitations

The primary limitation is that the 30-day auto-review requires an active cron integration; if the agent is not initialized or the system is offline during the review window, the audit may be delayed. For best results, ensure your decision-log file names are descriptive slugs to facilitate easier searching in your memory folder. Avoid storing sensitive credentials or PII inside these logs, as the files are plaintext markdown. Always provide clear, bulleted lists for 'Pros' and 'Cons' to help the agent’s future analysis algorithms categorize the decision outcomes more accurately.