v1.0.0

Safe Long-Run Mode GPT5.4

BenignClawScan verdict for this skill. Analyzed May 1, 2026, 8:26 AM.

Analysis

This is a coherent instruction-only workflow for safer long-running GPT-5.4 tasks, with some normal cautions around checkpoints, subagents, and external services.

GuidanceThis skill appears safe to install as an instruction-only workflow. Before using it on sensitive work, decide where checkpoints should be saved, what data subagents may receive, and which external-service actions require your explicit approval.

Findings (3)

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.

Abnormal behavior control

Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.

Tool Misuse and Exploitation

SeverityLowConfidenceHighStatusNote

SKILL.md

Be careful with: Azure / Microsoft Graph; ClawHub / GitHub-backed operations; Orgo runtime and VM usage; websites / browser automation; messaging providers

The workflow anticipates use of external systems where existing agent permissions could cause real changes, although it frames this as cautionary and does not add new tools or credentials.

User impactIf used during a task that already has access to external services, the agent may perform long-running or batched operations against those systems.

RecommendationSet clear per-task limits, require confirmation for writes or public/account-impacting actions, and prefer dry runs for external-service changes.

Sensitive data protection

Checks for exposed credentials, poisoned memory or context, unclear communication boundaries, or sensitive data that could leave the user's control.

Memory and Context Poisoning

SeverityLowConfidenceHighStatusNote

SKILL.md

Always leave artifacts that make recovery easy: - notes - drafts - partial outputs - checkpoint files - project updates - result summaries

The skill intentionally creates persistent recovery artifacts, which is useful for long tasks but can retain sensitive task details if users do not scope what is saved.

User impactPrivate project details, intermediate work, or sensitive notes could remain in checkpoint files after the task ends.

RecommendationChoose a known workspace location for checkpoints, avoid saving secrets or credentials, and clean up recovery artifacts when they are no longer needed.

Insecure Inter-Agent Communication

SeverityLowConfidenceHighStatusNote

SKILL.md

Use subagents as workers ... Delegate when: tasks are independent; multiple files or systems are involved; work may take a while

The skill encourages delegation to subagents for long or parallel work, which may share task context across agents even though no specific agent channel or credential is introduced.

User impactDelegated agents may receive files, context, or task details that the user expected to stay in the main conversation.

RecommendationLimit delegated context to what each subtask needs and require approval before subagents perform sensitive external actions.