Reported issues for pyslang-mcp

Pod holds 6 of 6 problems reported by people outside the maintainer team. Issues filed by the project's own owners, members and collaborators are excluded entirely — a maintainer's release checklist is not a warning to a prospective user.

Back to pyslang-mcp.

Most discussed

Improve pyslang-mcp Support For Generated Verilog Analysis

Summary

While using pyslang-mcp to analyze a generated Verilog cache from an RTL frontend, the existing tools were useful for project parsing and design-unit summaries, but several common RTL-debug queries still required falling back to text search. The biggest gaps were local signal/member lookup, assignment extraction, and connectivity tracing across instance port bindings.

This issue requests additional MCP functions that would make generated-Verilog and large-cache analysis much more

Read the thread · 2026-06-08 · closed · 1 comment

Full-project analysis crashes on unconnected ports (reported as generic "Analysis failed")

What happens

Running any analysis tool (get_diagnostics, get_project_summary, parse_filelist) over a real multi-module design fails with the generic:

Analysis failed while running the tool.

parse_files on the same files individually works fine, which made it look like an environment or filelist problem. It isn't — it's a crash inside the hierarchy indexing that server.py swallows into that opaque message.

Minimal repro

Any design with an open port trips it. This is e

Read the thread · 2026-07-21 · open · external user · 0 comments

Relax local filesystem scope; focus security on non-exfiltration

Problem

The current single-project_root containment model is too intrusive for real production HDL environments.

Designers often need to analyze or compare code across multiple company-controlled filesystem locations, for example:

• current block checkout
• previous release snapshot
• shared IP repository
• generated or centralized filelist area
• internal NFS/farm paths
• local sibling workspaces used for comparison or review

In many companies, reading those local/internal paths is nor

Read the thread · 2026-06-04 · open · 0 comments

Explain why pyslang-mcp is read-only

Please explain in the README why pyslang-mcp is intentionally read-only.

The explanation should cover:

• why the read-only boundary is a security and IP-protection choice
• why pyslang-mcp is a compiler-backed reader/checker rather than an RTL-writing tool
• why the LLM is used for generating/editing RTL while pyslang-mcp is used to verify and explain existing HDL

This will help set expectations for users who might otherwise expect MCP tooling to edit source directly.

Read the thread · 2026-06-03 · closed · 0 comments

Add a security test plan for pyslang-mcp

Track implementation of a security test plan for pyslang-mcp.

Draft plan:

• docs/superpowers/plans/2026-06-03-pyslang-mcp-security-test-plan.md

Scope:

• Project-root containment and filelist-abuse tests
• Local transport/auth hardening, including internal HTTP and metadata/SSRF checks
• Output hygiene, truncation, cache isolation, and no-traceback assertions
• Offline egress validation, secret scanning, dependency/container checks, and parser fuzzing
• Security-specific skill evals for exf

Read the thread · 2026-06-03 · closed · 0 comments

Document broader RTL skill integration

Summary

Document how users with broader RTL analysis or audit skills can delegate Verilog/SystemVerilog compiler evidence gathering to pyslang-verilog-context and pyslang-mcp.

Motivation

Users may already have umbrella skills such as rtl-analysis, rtl-audit, or llm-based-verilog-analysis. The README should explain how to compose those broad skills with pyslang-verilog-context so structural claims, diagnostics, hierarchy, symbol lookup, includes, defines, and filelist behavior

Read the thread · 2026-05-27 · closed · 0 comments

Most recent

The remaining reports are on the project's issue tracker.