Compress checklist command template (27% fewer tokens, same semantics)

[Xopoko]

Description

Proof of concept for #2943: editorial compression of templates/commands/checklist.md — the largest command template — from 4,460 to 3,257 tokens (-27%, o200k_base). The template body is rendered into every agent integration and loaded into the model's context on every /speckit.checklist invocation, so this is a per-use saving for every Spec Kit user.

This is a wording change, not a behavior change. Invariants held (mechanically checked against origin/main):

• frontmatter byte-identical;
• every fenced block containing commands or literal output templates byte-identical — including both extension-hook EXECUTE_COMMAND: blocks;
• all placeholder tokens and .specify/ paths preserved;
• every numbered step and MUST/NEVER/"do not" rule preserved with full semantics; only boilerplate prose, restatements, and surplus examples were cut (at least one example per concept retained);
• rule lists keep one-rule-per-bullet structure (restored after the Copilot review — compression comes from prose removal, not from collapsing structure).

The full per-template methodology, measurements for all nine templates, and the discussion this PR is a starting point for live in #2943. The remaining eight templates are ready on chore/2943-compress-command-templates and I'm happy to submit them in whatever granularity the maintainers prefer.

Testing

• Tested locally with uv run specify --help
• Ran existing tests with uv sync && uv run pytest (all pass)
• Tested with a sample project: uv run specify init --integration claude from this branch renders all skills correctly (rendered output diffed against the template — only the expected {SCRIPT}/path substitutions differ); we also run the compressed templates in a production multi-repo SDD workspace where a full specify→plan→tasks→implement cycle behaved identically to the verbose originals

AI Disclosure

• I did not use AI assistance for this contribution
• I did use AI assistance (describe below)

The compression draft was produced with LLM assistance under the hard invariants above, then verified by a deterministic invariant script (frontmatter equality; placeholder, path, and fenced-block inventory diff) and an adversarial line-by-line semantic review; violations found by review were restored before submission. I personally reviewed and tested the result (pytest, rendered init, and the production workspace run described above).

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR streamlines the checklist command template by collapsing many multi-line guideline sections into shorter, more compact paragraphs, while emphasizing that checklists validate requirement quality (not implementation behavior).

Changes:

• Condensed introductory “unit tests for English” explanation and examples.
• Simplified extension hook instructions and clarifying-question guidance into fewer, denser bullets/paragraphs.
• Compressed checklist generation rules (naming, append behavior, item structure) and domain examples.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Xopoko]

Addressed the review in 509bba5: all collapsed rule chains are back to one-rule-per-bullet structure (labeled NOT/FOR blocks, question rules, context-loading strategy, file create/append/preserve rules, quality dimensions, item structure). That was the right call — the structure costs only 74 tokens. Updated numbers: 4,460 → 3,257 tokens (-27%). Invariant checks still pass; same restructuring mirrored on the full nine-template branch.