sync: port triage agent changes from scaffold

[ggallen]

Summary

Sync triage agent files with fullsend-ai/fullsend scaffold, bringing the agents repo up to date with 17 commits that landed in the scaffold since the initial fork.

Changes by commit:

1. Schema: Replace blocked action (single blocked_by URL) with prerequisites (structured existing/create arrays); add question action
2. Agent prompt: Prerequisites action, existing-PR gate (hard constraint), question action, scoped anti-premature-resolution rule
3. Post-script: Prerequisites handler with cross-repo issue creation and allowlist validation, question handler, stderr diagnostics, deferred ready-to-code label, GITHUB_WORKSPACE guard
4. Pre-script: Add question label to triage label reset set
5. Tests: Full test coverage for prerequisites and question actions
6. Skill: Generalize issue-labels for issues and PRs (not just issues)
7. Harness: Migrate runner_env → env.runner/env.sandbox per ADR 0055

Original commits ported:

• 0fb3793c fix(#1752): defer ready-to-code label until after label_actions
• 1e725001 feat(#1811): add question action to triage agent
• 1a2db28a fix: update stale doc references for question control label
• e492ac78 feat(schema): replace blocked with prerequisites action
• b2055cb1 feat(triage): replace blocked action with prerequisites in agent prompt
• 6f79d87a fix(triage): correct label name in agent prompt
• 3a44b0cc feat(triage): handle prerequisites action in post-script
• 080368cf fix(triage): update post-triage tests for prerequisites action
• 11bae491 fix(triage): update schema validation tests for prerequisites action
• 2933555f fix(#2024): triage agent must not ask reporters implementation questions
• 8c7bec88 fix(#2237): scope anti-premature-resolution rule to user-facing questions
• e57f10a7 fix(triage): address review feedback on prerequisites action
• 12b47a9a fix(#2393): add diagnostic stderr output to post-script failure paths
• 095039eb fix(#1321): add existing-PR gate to triage agent definition
• 3ed6080c feat(skill): generalize issue-labels for issues and PRs
• 51ec19e2 fix(#2458): guard GITHUB_WORKSPACE in post-triage.sh for local execution
• 3ce10f4d refactor(harness): migrate triage to env.runner/env.sandbox (ADR 0055)

Test plan

• post-triage-test.sh — all 46 tests pass
• validate-output-schema-test.sh — all triage-specific tests pass (non-triage schema tests expected to fail since fix/review schemas not in this repo)

[ggallen]

Closing — took the wrong approach by overwriting agents repo files with scaffold versions. The agents repo files are already up to date; the scaffold is the one that needs to catch up or be deprecated.

[fullsend-ai-review]

🤖 Finished Review · ✅ Success · Started 2:44 PM UTC · Completed 2:51 PM UTC
Commit: 97a323f · View workflow run →

[fullsend-ai-retro]

🤖 Finished Retro · ✅ Success · Started 2:49 PM UTC · Completed 2:56 PM UTC
Commit: 97a323f · View workflow run →

[fullsend-ai-review]

Review skipped — this PR is already merged.

The /fs-review command only reviews open pull requests.

_{Posted by fullsend post-review check}

[fullsend-ai-retro]

Retro: PR #8 — sync: port triage agent changes from scaffold

Timeline

1. 14:41:11 — PR opened by ggallen (manual sync, branch sync-triage-from-scaffold). 15 files, 2,246 lines added — ports the full triage agent system from scaffold.
2. 14:41:25 — Review agent dispatched (run 28525802151).
3. 14:41:37 — Human comments that the PR took the wrong approach (overwriting agents repo files with scaffold versions).
4. 14:41:38 — PR closed.
5. 14:41:49 — PR reopened (11 seconds later).
6. 14:44:18 — Review bot posts status comment (review started on 97a323f).
7. 14:44:38 — PR merged by ggallen (~3.5 minutes after open).
8. 14:51:59 — Review agent completes (~10.5 minutes total, 31 tool calls, 4 sub-agents). Post-script detects PR is already merged and skips posting the review.
9. 14:49:44 — Retro agent dispatched.

Analysis

Review timing: The review agent did substantial work (31 tool calls, 4 sub-agents, ~7 minutes of agent time) but its output was never posted because the PR merged before the review completed. This is a known class of issue.

No review before merge: The PR was merged without any formal review — no human approval, no bot review. For a manual sync PR of this size, this is a process choice by the maintainer.

Retro debounce worked correctly: The first retro dispatch (on PR reopen) was properly debounced/cancelled, and only the post-merge retro ran.

Existing coverage

All identified improvement areas are already tracked by open issues in fullsend-ai/fullsend:

• Review agent wasted on merged PRs: #2388, #2606, #1439
• Post review results on merged PRs as informational: #1373
• Pre-review skip to save tokens: #2809

No new proposals are warranted — existing issues comprehensively cover the improvement opportunities observed in this workflow.