Security audit
Security checks for vulnerabilities and agentic risk
The skill is a log-compression helper, but it asks for execution and memory-writing authority while leaving the actual script and data-flow boundaries unclear.
Install only if you are comfortable letting the agent read log files, run a local compression command, and append reviewed summaries to MEMORY.md. Before using it, confirm what script is actually being executed and whether any log content is sent to an LLM or remote service.
The skill makes contradictory trust claims: it says it is a purely local Node.js script with no external dependencies, yet elsewhere states an LLM API is required. This can mislead users and agents about the actual execution and data-flow boundary, causing sensitive logs to be processed under false assumptions about locality and privacy.
The workflow instructs writing output to /tmp and appending it into MEMORY.md without a prominent warning or explicit confirmation step before modifying files. Because the skill also requests exec/write capabilities and operates on long-term memory artifacts, this raises the risk of unintended persistence, corruption, or poisoning of future agent context.
The example trigger "今天的日志太长了，帮我压缩" relies on the broad phrase "帮我压缩" rather than a narrowly scoped invocation pattern. In a markdown skill description, this can make activation boundaries unclear and increase the chance of unintended invocation during ordinary conversation.
The documentation again conflicts with earlier assurances by declaring an LLM API as a required dependency despite claiming the tool is local and requires no API key. For a log-compression skill handling potentially sensitive memory files, ambiguity about whether content leaves the local environment is a meaningful security and privacy risk.
No suspicious patterns detected.
