lsp-impact

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references a dependency on the agent-lsp MCP server hosted at github.com/blackwell-systems/agent-lsp. This repository belongs to the skill's author and serves as a legitimate technical requirement.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes user-supplied file paths and symbol names which are interpolated into tool parameters. Additionally, it reads source code files that could theoretically contain adversarial instructions designed to influence the agent's reasoning during the analysis phase.
  • [DATA_EXFILTRATION]: The skill uses tools to traverse the local filesystem and extract symbol relationships, call hierarchies, and reference counts. While this provides the agent with access to technical project data, the operations are explicitly read-only and the skill lacks any network-based exfiltration mechanisms to send this data to external domains.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 07:28 AM
Security Audit — agent-trust-hub — lsp-impact