refactor(mcp): describegpt llm passthru tweaks

[jqnatividad]

No description provided.

[Copilot]

Pull request overview

Refactors the MCP integration around describegpt to better support MCP sampling and an “agent-as-LLM” fallback path, while tightening which CLI flags/examples are exposed in MCP-generated skill definitions and updating related docs/tests.

Changes:

• Adjust MCP skill generation to omit describegpt examples and strip describegpt CLI flags that don’t apply in MCP mode.
• Extend MCP server/tooling to treat describegpt as non-tabular output, add a fallback result message when stdout is empty, and add an agent-driven two-call flow when MCP sampling isn’t available.
• Update docs/tests and the qsv-describegpt skill JSON to reflect MCP sampling options and the expanded COMMON_COMMANDS list.

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
src/mcp_skills_gen.rs	Skips describegpt examples and filters MCP-irrelevant describegpt options during skill generation.
src/help_markdown_gen.rs	Preserves MCP as an uppercase acronym in generated headings.
docs/help/describegpt.md	Documents --prepare-context / --process-response under “MCP Sampling Options” and updates usage/nav.
.claude/skills/tests/mcp-server.test.ts	Updates COMMON_COMMANDS count expectation to 12.
.claude/skills/tests/deferred-loading.test.ts	Updates COMMON_COMMANDS count expectation to 12.
.claude/skills/src/utils.ts	Adds findOutputPath + describegptFallbackResult helpers.
.claude/skills/src/mcp-tools.ts	Adds describegpt passthrough logic, sampling/fallback orchestration, stricter param handling, and auto output file naming.
.claude/skills/src/mcp-server.ts	Wires MCP server through to generic tool handling.
.claude/skills/src/mcp-sampling.ts	Exports runQsvCapture and uses describegptFallbackResult when stdout is empty.
.claude/skills/qsv/qsv-describegpt.json	Removes MCP-irrelevant options/examples from the describegpt skill definition.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 4 comments.

Comments suppressed due to low confidence (1)

.claude/skills/src/mcp-tools.ts:2592

• buildDescribegptArgs iterates over all params and will emit duplicate flags/values if the same option is provided in multiple key forms (e.g., "dictionary" and "--dictionary"). This can lead to ambiguous precedence when values differ; consider tracking seen normalized keys and choosing a single source of truth (e.g., prefer canonical keys over dashed keys).

  for (const [rawParam, value] of Object.entries(params)) {
    // Normalize: strip leading --, convert - to _
    const normalized = rawParam.replace(/^--/, "").replace(/-/g, "_");
    const flag = reverseLookup.get(normalized);
    if (!flag) continue;
    if (value === undefined || value === null || value === false) continue;
    if (value === true) {
      args.push(flag);
    } else {
      args.push(flag, String(value));
    }
  }