Superpowers Dispatching Parallel Agents
Security checks across static analysis, malware telemetry, and agentic risk
Overview
This instruction-only skill is coherent and purpose-aligned, but users should supervise parallel subagents and limit the context they share.
This skill appears safe to install as an instruction-only workflow aid. Before using it, make sure the tasks are independent, give each subagent minimal context, and review all returned changes before integrating them.
Static analysis
No static analysis findings were reported for this release.
VirusTotal
VirusTotal findings are pending for this skill version.
Risk analysis
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.
Subagents may receive task details or excerpts from the main session, so unnecessary private or sensitive context could be exposed to additional agent contexts.
The skill explicitly uses subagents and context handoff. It also recommends isolation, which is a good control, but users should still be careful about what information is copied into subagent prompts.
You delegate tasks to specialized agents with isolated context... They should never inherit your session's context or history — you construct exactly what they need.
Give each subagent only the files, error messages, and constraints it needs; avoid including unrelated secrets, credentials, or private session history.
Parallel work can save time, but conflicting or inconsistent changes may be introduced if the results are merged without review.
Concurrent subagents can make independent changes that later interact or conflict. The skill acknowledges this by requiring review, conflict checks, and full-suite verification.
Spawn subagents simultaneously — each handles one domain... When agents return: Read each summary; Verify fixes don't conflict; Run full test suite
Use this only for truly independent tasks, review each subagent’s changes, check for overlapping edits, and run full verification before accepting the combined result.