Add closed-vocabulary preset to community catalog

[yunusdim]

Adds closed-vocabulary to the community catalog. Requested in #4106.

What it does. Appends one pass to /speckit.analyze that flags closed sets of values enumerated more than once with different members, and reports its own coverage on every run.

Preset: https://github.com/yunusdim/spec-kit-preset-closed-vocabulary — MIT, tagged v1.0.1.

Uses strategy: "append", so it carries no copy of templates/commands/analyze.md and does not need to track upstream changes to that file. Requires >=0.8.0 because composition strategies landed there (#2133).

Evidence. Three fixtures ship with the preset under evidence/fixtures/. One must fire, two are negative controls — a formatting-variance control and a declared-subset control. evidence/RESULTS.md separates the reproducible bench from a field measurement taken on a private corpus, and labels the second as not verifiable by anyone else.

Composition was verified on device: the composed file carries the core untouched (lines 1-268) with the pass appended after it, confirmed by direct inspection of the composed file rather than assumed from the append strategy. Detection was verified mechanically: the rules in commands/speckit.analyze.md were implemented as a standalone, read-only instrument and run against the three fixtures shipped in this repository. All three matched — 001-divergent fired, 002-agreeing and 003-declared-subset did not. Building that instrument caught a real bug: an unrecognised set-membership introducer produced a false negative on 001-divergent (coverage 2 of 26 lines). Fixed before tagging; the run in evidence/RESULTS.md is the corrected one. Declared limit: this is composition and mechanical rule verification, not an agentic end-to-end invocation of /speckit.analyze — whether a model reading the composed prose applies the same rules identically is not established by this evidence.

Coverage is reported unconditionally, including when nothing is found. A zero-divergence result without a coverage figure says nothing about the spec — it says something about what the pass managed to read. The reference implementation reported zero across a corpus that had a divergence in plain sight, because it failed to recognise one introducer and one separator, having parsed 4 per cent of what it inspected.

It warns. It never blocks.

Entry inserted alphabetically between claude-ask-questions and command-density. download_url matches the install command in the preset README character for character.

[Copilot]

Pull request overview

Adds the closed-vocabulary preset to the community catalog.

Changes:

• Registers preset metadata, compatibility requirements, capabilities, and release links.
• Places the entry alphabetically in the catalog.

Show a summary per file

File	Description
presets/catalog.community.json	Adds the closed-vocabulary catalog entry.

Review details

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

• Files reviewed: 1/1 changed files
• Comments generated: 2
• Review effort level: Balanced

[mnriem]

@yunusdim Can you create a new preset submission issue for this? This will allow us to run our automated agentic workflows. Thanks!

[yunusdim]

@mnriem Done — #4192.