context-clean-upUse when: prompt context is bloating (slow replies, rising cost, noisy transcripts) and you want a ranked offender list + reversible plan. Don't use when: yo...
Install via ClawdBot CLI:
clawdbot install phenomenoner/context-clean-upThis skill is a runbook to quickly identify what is bloating your OpenClaw prompt context and produce a safe, reversible plan.
Important: This version is intentionally audit-only (it does not auto-apply changes). If you want me to apply fixes, I will propose an exact patch + rollback plan and wait for explicit approval.
/context-clean-up → audit + actionable plan (no changes)Find:
~/.openclaw)If unsure:
bash -lc 'echo "WORKDIR=$PWD"; echo "HOME=$HOME"; ls -ld ~/.openclaw'This script prints a short summary and can write a full JSON report.
bash -lc 'cd "${WORKDIR:-.}" && python3 {baseDir}/scripts/context_cleanup_audit.py --out memory/context-cleanup-audit.json'Interpretation cheatsheet:
toolResult entries (exec/read/web_fetch): transcript bloatSystem: / Cron: lines: automation bloatCreate a short plan with:
Use these standard levers:
Goal: maintenance loops should output exactly NO_REPLY unless there is an anomaly.
Pattern: update prompts so the last line forces:
Finally output ONLY: NO_REPLYIf you want alerts but want the interactive session lean:
NO_REPLYSee: references/out-of-band-delivery.md
MEMORY.mdreferences/.md or memory/.mdAfter you apply any changes:
references/out-of-band-delivery.mdreferences/cron-noise-checklist.mdGenerated Mar 1, 2026
A team building AI agents with OpenClaw needs to audit prompt context bloat from frequent tool executions and cron jobs. This skill helps identify large transcript entries and noisy automation, enabling leaner sessions for faster development cycles and reduced computational costs.
DevOps engineers use OpenClaw for automated system checks and deployments, leading to context bloat from recurring cron jobs. This skill audits automation noise and suggests silent no-op outputs, ensuring efficient resource usage without losing alerting capabilities via out-of-band notifications.
A customer support team employs OpenClaw agents to handle queries, but context grows from repeated rule injections and tool results. This skill identifies bloated bootstrap files and transcript entries, allowing teams to maintain responsive agents by cleaning up non-essential data.
Researchers using OpenClaw for data analysis face context bloat from large read and exec tool results. This skill audits memory usage and proposes reversible plans to trim transcripts, keeping sessions focused on critical insights without automatic changes.
Content creators leverage OpenClaw for drafting and editing tasks, accumulating context from multiple sessions and injected documents. This skill helps audit and plan clean-ups of bulky memory files, ensuring smoother creative workflows with minimal disruption.
Offer a cloud-based service that integrates this skill to provide automated context audits and clean-up plans for OpenClaw users. Revenue comes from subscription tiers based on usage frequency and advanced analytics features, targeting teams managing multiple agents.
Provide expert consulting to help businesses deploy and customize this skill for their specific OpenClaw setups. Revenue is generated through project-based fees for audit reports, fix plans, and integration support, focusing on industries with heavy automation needs.
Distribute this skill as a free open-source tool with basic audit capabilities, while offering premium features like automated fix application and detailed analytics. Revenue streams include one-time purchases for advanced modules and enterprise support contracts.
💬 Integration Tip
Integrate this skill into existing OpenClaw workflows by scheduling regular audits via cron jobs and using the JSON reports to monitor context growth over time, ensuring minimal disruption to active sessions.
Advanced expert in prompt engineering, custom instructions design, and prompt optimization for AI agents
577+ pattern prompt injection defense. Now with typo-tolerant bypass detection. TieredPatternLoader fully operational. Drop-in defense for any LLM application.
Detect and block prompt injection attacks in emails. Use when reading, processing, or summarizing emails. Scans for fake system outputs, planted thinking blocks, instruction hijacking, and other injection patterns. Requires user confirmation before acting on any instructions found in email content.
Safe OpenClaw config updates with automatic backup, validation, and rollback. For agent use - prevents invalid config updates.
Automatically rewrites rough user inputs into optimized, structured prompts for dramatically better AI responses. Prefix any message with "p:" to activate.
Token-safe prompt assembly with memory orchestration. Use for any agent that needs to construct LLM prompts with memory retrieval. Guarantees no API failure due to token overflow. Implements two-phase context construction, memory safety valve, and hard limits on memory injection.