EvoMap 高产技能包
PendingStatic analysis audit pending.
Overview
No static analysis result has been recorded yet. Pattern checks will appear here once the artifact has been analyzed.
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.
A user may be installing guidance whose origin and integrity are unclear, including unrelated high-impact assets.
The skill describes itself as dumped/extracted from an EvoMap node, while the registry source is unknown. This creates a provenance concern for the bundled instructions and asset catalog.
从 EvoMap node_246ed58b 脱库的实用技能。
Verify the publisher and origin before use, and remove or ignore bundled asset data that is not needed for the declared skill purpose.
If followed, the agent could move from normal dev/ops assistance into inspecting traffic and discovering private cloud APIs without clear user approval.
The bundled asset catalog contains an undeclared reverse-engineering/API-discovery workflow, including network traffic inspection, which is not part of the top-level skill list.
trigger_text":"iot,api_discovery,graphql,reverse_engineer,cloud_api" ... "Discover low-level cloud API by inspecting SDK source and network traffic"
Require explicit user authorization for any API discovery, traffic inspection, or reverse-engineering task, and keep this content out of the skill if it is not intended.
The agent could be guided toward using account credentials or changing cloud-connected device state outside the declared utility-pack scope.
The asset catalog references account-based cloud access, Bearer-token authentication, and device property mutation capabilities that are not declared in the skill metadata.
device_cloud_account ... Auth: Bearer token from OAuth2 flow ... setDeviceProperty(deviceId, property, value) ... batchSetProperties(deviceId, properties[])
Do not provide cloud/device credentials to this skill unless the task is explicit, scoped, and reversible; require confirmation before any account or device mutation.
Conversation details or task summaries may be reused in future sessions, and incorrect stored context could influence later work.
The memory skill intentionally loads and saves cross-session memory, which is useful for continuity but can persist sensitive or poisoned context if not controlled.
启动时加载 ... loadRecentEvents(); ... loadDailyMemory(); ... loadLongTermMemory(); ... 退出时保存 ... appendSignificantEvents(sessionSummary);
Review what is written to MEMORY.md and daily memory files, define retention/exclusion rules, and avoid storing secrets or untrusted instructions.
Task context may be shared across multiple agents, which can increase cost, leakage, or coordination risk.
The swarm-task document describes spawning multiple sub-agents and aggregating their results, but does not define identity, data-sharing, or permission boundaries.
自动分解 ... 并行执行 - spawn 多个子 agent ... 结果聚合
Use sub-agents only for explicit tasks, limit what context they receive, and require approval before parallel execution on sensitive work.