メニュー
Run full LSP validation on a Tcl file and produce a categorised report of all issues: errors, security, style, and optimiser suggestions. Use when validating Tcl code, linting .tcl files, checking Tcl script quality, or running static analysis on Tcl scripts.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Compare our compiler's bytecode output against tclsh reference disassembly (8.4, 8.5, 8.6, 9.0). Use when verifying bytecode correctness, investigating instruction sequence differences, or checking that our codegen matches C Tcl.
Debug the Tcl compiler pipeline (lexer → green tree → segmenter → IR → CFG → SSA → codegen) running from source — never a built pyz/zipapp. Wraps the tcl-explorer CLI and adds source-slice and raw-token views that make off-by-one range bugs obvious. Use when investigating a wrong diagnostic position, a stray-brace / unmatched-delimiter false positive, a mis-tokenised word, an IR statement whose range looks wrong, or any time you need to see what the compiler produces for a snippet of Tcl.
Download and extract Tcl source trees (8.4, 8.5, 8.6, 9.0) to tmp/ for test suites and bytecode reference. Idempotent — skips versions already present.
Manage differential-fuzzer findings: query fixed/unfixed status, list by category, verify findings against the current VM, mark findings as fixed/unfixed, and batch-update categories. Use when working on fuzz findings, checking what's fixed, or triaging new results.
Use when verifying Tcl LSP server behavior: semantic tokens, diagnostics, formatting, hover, completion, definition, references, code actions, optimizations, document symbols, diagram extraction, event/command registry lookups, or benchmarking server performance and collecting timing logs. Tests the server directly over JSON-RPC without VS Code.
Query reference test results captured from real tclsh (8.4–9.0). Look up individual tests, search groups, view suite summaries, and compare results across Tcl versions.
| name | tcl-validate |
| description | Run full LSP validation on a Tcl file and produce a categorised report of all issues: errors, security, style, and optimiser suggestions. Use when validating Tcl code, linting .tcl files, checking Tcl script quality, or running static analysis on Tcl scripts. |
| allowed-tools | Bash, Read |
Run full validation on a Tcl file and produce a categorised diagnostic report.
ai/prompts/tcl_system.mduv run --no-dev python ai/claude/tcl_ai.py validate $FILE
See docs/generated/diagnostic_codes.md for the full auto-generated table of all diagnostic codes with descriptions and defaults.
Use headings for each category. For each diagnostic, show:
Example structure:
### Errors (1)
- **E001** (line 12): Missing subcommand — add the required subcommand after `string`
### Style (1)
- **W100** (line 5): Unbraced expression — wrap in braces for bytecode compilation
### Summary
- Errors: 1, Style: 1, Total: 2
$ARGUMENTS