multi task
PassAudited by ClawScan on May 1, 2026.
Overview
This instruction-only skill is aligned with batch processing, but it can fan work out to multiple subagents and copy task context into their prompts, so users should confirm the scope before use.
This skill looks coherent for speeding up independent batch work. Before using it, review the list of files or tasks, confirm the output locations, avoid sending secrets as shared context, and ask for a small pilot run before approving large or sensitive batches.
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.
Private project details or file contents may be repeated into several subagent prompts while processing a batch.
The skill intentionally copies context into prompts sent to multiple subagents. This is necessary for its purpose, but it can spread sensitive user or file context beyond the main conversation.
Each subagent starts with a blank context — it cannot see this conversation. Every prompt must be completely self-contained... Shared context is included verbatim, not by reference
Before dispatching, review the shared context and avoid including secrets or unrelated sensitive information in subagent prompts.
A bad batch plan could create many incorrect outputs or repeated errors faster than serial execution.
Parallel task dispatch is the core feature. If the plan or prompt is wrong, the same mistake could be applied across many files or outputs quickly, although the skill includes planning and confirmation guidance.
The core insight: a single message can contain multiple Task tool calls, and they all execute concurrently.
Confirm the plan, use pilot tasks for large or high-impact batches, and keep outputs isolated from original files.
Long-running background task execution may continue consuming resources or performing batch work after the user expects the interaction to be complete.
The reference guidance mentions background execution for long-running batches. It appears purpose-aligned, but background work should have clear user approval and stopping conditions.
If tasks are taking very long → keep size but add background execution
Use background execution only with explicit approval, visible status updates, and clear stop or completion criteria.