docs(#1662): ADR 0043 — require authorization on all slash commands

[fullsend-ai-coder]

Add ADR proposing that /fs-triage, /fs-code, and /fs-review use the same is_authorized gate already enforced by /fs-fix, /fs-retro, and /fs-prioritize. The ADR addresses the four design questions from the issue: automatic event triggers remain ungated, bot-to-bot workflows are preserved via the existing Bot-type bypass, unauthorized users see silent failure (consistent with existing gated commands), and is_authorized is a platform-level boundary not overridable per-repo.

Note: make lint could not run due to sandbox Go toolchain permission error. ADR-specific linters (lint-adr-frontmatter, lint-adr-numbers, lint-adr-status) all passed.

---

Closes #1662

Post-script verification

• Branch is not main/master (agent/1662-adr-auth-slash-commands)
• Secret scan passed (gitleaks — 11cdd7674a31c564785e9daa2efaf7b931385e1c..HEAD)
• Pre-commit hooks passed (authoritative run on runner)
• Tests ran inside sandbox

[github-actions]

Site preview

Preview: https://8eefdf73-site.fullsend-ai.workers.dev

Commit: dca81386536c432d2ab4c12467c83211b0ea7cf4

[fullsend-ai-review]

Review

Findings

No findings.

The ADR accurately describes the current authorization gap — verified against reusable-dispatch.yml that /fs-triage, /fs-code, and /fs-review lack the is_authorized gate present on /fs-fix, /fs-retro, and /fs-prioritize. The proposed solution is consistent and addresses all four design questions from #1662: automatic triggers are gated by actor association, bot-to-bot workflows are preserved via the existing COMMENT_USER_TYPE != "Bot" bypass, unauthorized users receive visible feedback, and is_authorized is a non-overridable platform boundary. ADR number 0043 is sequential with no conflicts.

Previous run

Review

Findings

No findings.

The ADR is well-structured, addresses all four design questions from the linked issue (#1662), and proposes a consistent authorization pattern across all slash commands. Cross-references to ADR 0034, ADR 0042, and the security threat model are valid. The automatic-event and bot-bypass carve-outs are clearly scoped and well-reasoned. No existing documentation is made stale by this change.

Previous run (2)

Review

Findings

No findings.

The ADR is well-structured, factually accurate, and fully addresses all four design questions from #1662 (automatic event triggers, bot-to-bot workflows, error messaging, per-repo configurability). The authorization gating table was verified against reusable-dispatch.yml — the claimed gated/ungated status for each slash command is correct. Cross-references to ADR 0034 and ADR 0042 are valid. Frontmatter and section format are consistent with existing ADRs.