chore(#537): add GitHub issue templates with type labels

[fullsend-ai-coder]

Summary

Add GitHub issue forms for bug reports and feature requests so new issues arrive with structured prompts and initial type/bug / type/feature labels.

Blank issues stay enabled for other topics. The outcome-tracked template tracked in #1380 is intentionally out of scope.

Changes

• Add .github/ISSUE_TEMPLATE/bug.yml (auto-label type/bug) with required current behavior, expected behavior, and reproduction steps
• Add .github/ISSUE_TEMPLATE/feature.yml (auto-label type/feature) with required gap, desired behavior, and motivation
• Point CONTRIBUTING.md at the issue template chooser

Testing

• YAML parses and matches GitHub's issue form schema (name, labels, unique field ids, required validations)
• Direct execution of applicable pre-commit hooks: check-yaml --unsafe, end-of-file-fixer, trailing-whitespace, detect-private-key, check-added-large-files, check-merge-conflict, mixed-line-ending, gitleaks, lychee
• No Go production code changed, so go test was not run

Checklist

• PR title follows Conventional Commits (correct type; no breaking change)
• Commits are not DCO-signed (autonomous agent session)
• I wrote this contribution myself and can explain all changes in it

---

Closes #537

Post-script verification

• Branch is not main/master (agent/537-github-issue-templates)
• Secret scan passed (gitleaks — d5f36921ac754705619f38c637ef692873809fbc..HEAD)
• PR body secret scan passed (gitleaks — no-git)

[fullsend-ai-review]

🤖 Review · ⚠️ Cancelled · Started 5:37 PM UTC · Ended 5:37 PM UTC

Commit: c2fb5a6 · View workflow run →

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 5:38 PM UTC · Completed 5:50 PM UTC

Commit: c2fb5a6 · View workflow run →

Runtime: pi · Model: sonnet → claude-sonnet-5 · Effort: high · Cost: $3.20

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fullsend-ai-review]

Risk Assessment: moderate (2/5)

Details

Low-complexity, docs/config-only PR from a bot author adding two new GitHub issue templates and a CONTRIBUTING.md pointer, with no security-sensitive files, no dependency or CI changes, and a scope that deliberately implements a sanctioned subset of an open-ended linked issue; the composite is pulled from the low end to a moderate 2 specifically because both new files fall under the protected .github/ path prefix.

[fullsend-ai-review]

Review

Findings

Medium

• [protected-path] .github/ISSUE_TEMPLATE/bug.yml, .github/ISSUE_TEMPLATE/feature.yml — This PR modifies files under a governance/infrastructure path (.github/) on the active protected-paths list. The PR links to issue Add GitHub issue templates with auto-labels #537 and its description explains the rationale for adding these issue forms, so context is sufficient — but human approval is always required for protected-path changes regardless of context.

---

Labels: PR adds GitHub issue form templates and updates contributor-facing documentation.

[ralphbean]

in retrospect, I think this is unnecessary. triage and other agentic tools (like the filing-issues skill) help close this gap.

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 8:16 PM UTC · Completed 8:22 PM UTC

Commit: c2fb5a6 · View workflow run →

Runtime: claude · Model: sonnet → claude-sonnet-5 · Effort: high · Cost: $1.21

[fullsend-ai-retro]

PR #7174 implemented issue #537 (GitHub issue templates) cleanly — the code agent's templates matched repo conventions and the review agent correctly flagged the protected-path (.github/) change for required human approval. The waste was upstream: the human ran /fs-triage on the 4-month-old, low-priority issue asking "is this still relevant?"; the triage agent reasoned about the repo's new post-filing labeling pipeline but still recommended proceeding and applied ready-to-code, which auto-dispatched the code agent and then review. The human then closed both PR and issue as "unnecessary in retrospect," citing triage automation and the filing-issues skill as already closing the gap — costing roughly $5.50 in code+review agent spend for work that was discarded. This is the second confirmed instance of /fs-triage auto-cascading into unwanted code dispatch: open issue #2834 (filed 2026-07-01, still unimplemented) already proposes making /fs-triage recommend ready-to-code rather than auto-applying it, based on an earlier near-identical incident (issue #2779/PR #2833). This run is fresh corroborating evidence that #2834 remains unaddressed and the waste pattern is recurring ~2 months later. Separately, I found a genuinely novel and unfiled gap: the triage agent's sufficiency/relevance checks (fullsend-ai/agents agents/triage.md) only check for prior overlapping PRs, issues, or fixes (steps 2b/2d, and the related closed pattern in agents#903) — there is no instruction to check whether an existing agentic skill or automation capability (like filing-issues) already substantially satisfies a requested capability. That gap is filed below as a new proposal, scoped narrowly to avoid overlapping #2834 or #903.

Proposals filed

• Triage relevance/sufficiency checks should weigh existing agentic skills/automation, not just prior PRs and fixes (in fullsend-ai/agents)