Close issue #2574: recommendation contradicts ADR 43 and targets relocated code

[fullsend-ai-retro]

What happened

Issue #2574 was created by the retro agent on 2026-06-23, proposing that provenance headers should be embedded in the shim template (internal/scaffold/fullsend-repo/templates/shim-workflow-call.yaml). PR #2578 attempted to implement this but was rejected for two reasons: (1) the recommended approach violated ADR 43 (templates must remain header-free), and (2) the code has moved to fullsend-ai/agents per ralphbean's review. The issue remains open with ready-to-code label, meaning another code agent run could pick it up and repeat the same wasted cycle.

What could go better

The issue should be closed to prevent another code agent from attempting the same implementation on the same stale path. The underlying problem (reconciliation bot output diverging from template) may still be valid but would need to be re-filed against fullsend-ai/agents with a corrected recommendation that respects ADR 43.

Confidence: HIGH — the PR that attempted to implement this issue was closed by a human maintainer who confirmed the code has moved. The issue's "preferred" option was architecturally invalid per an accepted ADR.

Proposed change

Close issue #2574 with a comment explaining: (1) the recommended approach (embedding headers in template) contradicts ADR 43, (2) the target code has moved to fullsend-ai/agents, and (3) if the reconciliation drift problem persists, a new issue should be filed in the correct repo with a deploy-time injection approach. Remove the ready-to-code label to prevent code agent dispatch.

Validation criteria

Issue #2574 is closed. No further code agent runs are dispatched against it. If the underlying reconciliation drift problem recurs, it surfaces as a new issue targeting the correct repo.

---

Generated by retro agent from #2578

[fullsend-ai-triage]

🤖 Finished Triage · ✅ Success · Started 6:58 PM UTC · Completed 7:01 PM UTC
Commit: e8381e3 · View workflow run →

[fullsend-ai-triage]

Triage Summary

Severity: Low · Category: Housekeeping

Problem

Issue #2574 remains open with the ready-to-code label despite its recommended approach being invalid on two fronts:

1. ADR 43 violation: The recommended approach (embedding provenance headers in the shim template) directly contradicts ADR 43, which requires templates to remain header-free with headers injected at deploy time.
2. Code relocation: The target code has moved from this repo to fullsend-ai/agents, as confirmed by ralphbean's review on PR fix(#2574): embed provenance headers in shim template #2578.

PR #2578 already attempted this implementation and was closed without merge after catching both issues.

Recommended Action

1. Close issue Reconciliation bot shim output diverges from template while claiming sync #2574 with a comment explaining the ADR 43 conflict and code relocation.
2. Remove ready-to-code label from Reconciliation bot shim output diverges from template while claiming sync #2574 to prevent another code agent dispatch.
3. If the underlying reconciliation drift problem persists, a new issue should be filed in fullsend-ai/agents using a deploy-time injection approach that respects ADR 43.

Validation

- Issue #2574 is closed and ready-to-code label removed
- No further code agent runs dispatch against #2574
- If reconciliation drift recurs, it surfaces as a new issue in fullsend-ai/agents

Related

• Reconciliation bot shim output diverges from template while claiming sync #2574 — the stale issue to close
• PR fix(#2574): embed provenance headers in shim template #2578 — rejected implementation attempt
• AGENTS.md: document that agent/skill content has moved to fullsend-ai/agents #3157 — documents agent content migration to fullsend-ai/agents
• Retro agent should consult ADRs and glossary before proposing changes #1545 — retro agent should consult ADRs (root cause of the bad recommendation)

[fullsend-ai-prioritize]

RICE Priority Score: 0.9

Score breakdown

Dimension	Score	Reasoning
Reach	0.5	This is an internal housekeeping issue affecting the development team and CI agents within a single org. It prevents wasted code-agent cycles on a stale issue (#2574) but does not directly affect end users or strategic customers.
Impact	0.5	Low impact — the consequence of inaction is another wasted agent cycle attempting an implementation that will be rejected (as PR #2578 already was). A workaround exists: human reviewers would catch and reject any duplicate PR. The fix is purely process hygiene.
Confidence	0.9	Very high confidence — the issue is precisely scoped (close one issue, remove one label), the PR rejection context is well-documented, and the triage agent confirmed the assessment. There is no ambiguity about what needs to happen.
Effort	0.25	Trivial — the entire task is closing issue #2574 with an explanatory comment and removing the ready-to-code label. No code changes, no design work, no testing required.

Formula: (0.5 x 0.5 x 0.9) / 0.25 = 0.9