Batch
PassAudited by VirusTotal on May 12, 2026.
Overview
Type: OpenClaw Skill Name: batch Version: 1.0.0 The skill bundle describes a robust batch processing mechanism with features like progress tracking, checkpointing, error handling with retries, partial failure recovery, and rollback for destructive operations. All instructions to the AI agent in SKILL.md, errors.md, and strategies.md are operational and align with the stated purpose of processing multiple items. While the agent is instructed to perform file operations (e.g., saving temporary files, potentially deleting/modifying files with confirmation and rollback) and API calls, these are necessary capabilities for its function. There is no evidence of intentional malicious behavior such as data exfiltration, unauthorized remote control, persistence, or prompt injection designed to subvert the agent's core purpose.
Findings (0)
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.
If used on the wrong item list, a batch task could affect many files or records at once.
The skill anticipates potentially high-impact bulk actions such as deletion, but it frames them as requiring explicit confirmation.
**Confirm destructive ops:** "This will delete 200 files. Proceed?"
Use the dry run, verify item counts and targets, and require explicit confirmation before bulk deletes or modifications.
Sensitive input data or partial results could remain in temporary checkpoint or failure files after the batch finishes.
Checkpoint and partial-result files can persist batch state, item identifiers, and possibly task data on disk.
"results_path": "/tmp/batch_abc123_partial.json"
Choose safe checkpoint locations, avoid storing secrets in batch state, and clean up temporary failure or partial-result files when no longer needed.
If sub-agent delegation is used, batch items may be distributed beyond the primary agent context.
The parallel strategy can involve sub-agents, which may share batch items across agent boundaries; the artifact limits concurrency but does not define data-sharing boundaries.
Max 5 concurrent sub-agents (avoid overload)
Only delegate appropriate items, and verify the permissions and data-handling behavior of any separate delegate skill before using it with sensitive data.