v1.4.2

Researching In Parallel

BenignClawScan verdict for this skill. Analyzed May 1, 2026, 6:18 AM.

Analysis

This is a coherent parallel-research skill, but it intentionally spawns multiple research agents and keeps workspace research artifacts for reuse.

GuidanceBefore installing, make sure you are comfortable with a workflow that spawns several sub-agents, uses web/PDF/browser retrieval, writes multiple files to a workspace, and may keep source extracts for reuse. Use a new empty workspace, verify important citations independently, and clean up retained artifacts when finished.

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

Requires sessions_spawn ... Requires web search, file read/write access to workspace for sub-agents. Browser and PDF tools strongly recommended

The skill gives sub-agents web/document retrieval and workspace file-write capabilities. These are significant tools, but they are disclosed and directly aligned with the research workflow.

User impactThe agent may browse/fetch documents, process PDFs, use a browser fallback, and write multiple research files in the chosen workspace.

RecommendationUse a dedicated empty workspace and only provide private files or URLs that you actually want included in the research run.

Rogue Agents

SeverityLowConfidenceHighStatusNote

skill-config.json

"maxConcurrent": 3, "params": { "cleanup": "keep", "runTimeoutSeconds": 1200, "thinking": "high", "agentId": "main" }

The configuration allows several sub-agent sessions to run concurrently for up to 20 minutes and keeps their results. This is not hidden, but it affects cost, duration, and retained artifacts.

User impactA run can consume meaningful model time/tokens and leave sub-agent outputs available after completion.

RecommendationConfirm the model choices and expected cost before spawning, and clean up retained sessions or workspace artifacts when they are no longer needed.

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

Supports iterative research — outputs from one run can be fed back as starting-point sources for the next.

The skill intentionally persists research outputs and can reuse them in later runs, so inaccurate, stale, or adversarial source material could influence future reports if reused uncritically.

User impactSaved reports or source extracts may shape later research iterations, including any mistakes or misleading material they contain.

RecommendationReview and curate saved sources before reusing them, and avoid treating prior AI-generated outputs as authoritative without verification.