Context Budgeting Pro
PassAudited by ClawScan on May 1, 2026.
Overview
This appears to be a purpose-aligned context-management skill, but it uses persistent memory checkpoints, periodic compaction instructions, and a small shell helper that users should review.
This skill looks reasonable for managing long OpenClaw sessions. Before installing, review the included shell script, check whether the hardcoded workspace path matches your setup, and remember that the skill may summarize and clear raw context during long sessions.
Findings (4)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Installing the skill may lead the agent to run a local script that interacts with OpenClaw sessions.
The included helper script invokes the OpenClaw CLI as part of session compaction. This is aligned with the skill's context-budgeting purpose, but it is still executable local automation.
openclaw sessions --active 1 > /dev/null
Review the script and confirm the OpenClaw command is appropriate for your environment before allowing the agent to run it.
Important task details may be summarized and reused later, so mistakes in the summary could affect future work.
The skill intentionally stores summarized task state in a persistent memory file. This is useful for context management, but persistent summaries can shape later agent behavior if inaccurate or incomplete.
Update `memory/hot/HOT_MEMORY.md` with: - **Status**: Current task progress. - **Key Decision**: Significant choices made. - **Next Step**: Immediate action required.
Periodically review HOT_MEMORY.md for accuracy, especially before relying on it for long-running or sensitive tasks.
The agent may discard or de-emphasize raw context after summarizing it, which could reduce fidelity in later steps.
The skill describes periodic compaction and cleanup after summarization. This is purpose-aligned, but if the summary is poor, later work may proceed without the original raw context.
Heartbeat (every 30m) acts as the Garbage Collector (GC): 1. Check `/status`. If Context > 80%, trigger the **Checkpointing** procedure. 2. Clear raw data ... once the summary is extracted.
Use this skill when context pressure matters, and ask the agent to preserve raw details when exact wording or full outputs are important.
It may be harder to verify who maintains the skill or whether this package is the intended version.
The package has limited provenance information, and the registry metadata also differs from the bundled _meta.json slug/owner. This does not show malicious behavior, but it is a provenance ambiguity.
Source: unknown; Homepage: none
Prefer skills from known sources, and compare the registry metadata with the bundled metadata before installing in sensitive environments.