docs(skill): default to full six-section template on issue creation (SMI-4344)

[wrsmith108]

Summary

• Instruction-layer fix: SKILL.md now defaults to the full six-section template (Context → Problem → Proposal → Acceptance Criteria → Verification → Out of Scope) when Claude creates a Linear issue. Brevity is opt-in via trigger phrases ("quick issue", "one-liner", "just the AC", "brief", "terse", "minimum", "short").
• buildIssueTemplate() returns the new six-section shape, so npm run ops -- create-issue --template prints the same skeleton SKILL.md teaches.
• docs/issue-template.md adds a fully populated Full example (~350-char body) so Claude has a pattern to match, not just a rule to follow. Example body passes validate-description --stdin at exit 0.

No validator, exit-code, or API changes. Same tiered-enforcement model as the MCP gate: instructions + retroactive lint-issues audit.

Closes SMI-4344.

Why

The AC validator at scripts/lib/issue-description.ts enforces a floor (MIN_BODY_CHARS=120, MIN_AC_ITEMS=2, placeholder rejection) but not a target depth. Issues created via the CLI or MCP routinely land near that floor — just enough to pass — because Claude Code's default tone is terse. Reviewers then lack the Context, Problem, Proposal, Verification, and risk notes that older plan-driven issues carried.

Recent evidence: SMI-4326 and SMI-4327 both passed validate-description --stdin at exit 0, but read as brief. This PR makes detail the default and terseness explicit.

Test plan

In-session verification (done)

• npm run build — typecheck + esbuild clean
• npm test — 49/49 tests green (incl. updated buildIssueTemplate assertions)
• npm run ops -- create-issue --template — prints the new six-section skeleton
• Full example body piped to validate-description --stdin → exit 0

Clean-context verification (must run in a fresh Claude Code session after merge)

AC #3 of SMI-4344 requires a test issue created from a vague prompt without the word "detailed". This can only be validated honestly outside the session that designed the fix. Protocol after merge:

• Start a fresh Claude Code session in this repo (no context from this PR)
• Prompt: "create an issue for [small concrete task]" — do NOT use the words "detailed", "verbose", "full", or "template"
• Assert the generated body contains ## Context, ## Proposal (or ## Problem), ## Acceptance Criteria, and ## Verification
• Delete the test issue

If Claude produces only Context + AC (the old floor-passing shape), iterate on SKILL.md bullet #1 phrasing.

Risks

1. Instruction drift — Claude ignores the new rule. Mitigations: (a) the inline template is the first thing Claude reads; (b) docs/issue-template.md has a populated Full example; (c) npm run lint-issues audits retroactively.
2. Users who wanted brevity get verbose issues. Mitigation: seven explicit trigger phrases for brevity opt-in. Keep the list short — synonyms dilute the rule.
3. Six sections feel ceremonial for trivial chores. Mitigation: depth rule explicitly permits collapsing Problem into Context and dropping Verification when AC is self-evidently testable — collapse deliberately, not by default.

Plan

Full plan at ~/.claude/plans/smi-4344-default-verbose.md (local to author; not checked in).

🤖 Generated with Claude Code

[github-actions]

🎉 This PR is included in version 3.2.0 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀