Add mandatory API report regeneration instruction to repo-wide agent config

[fullsend-ai-retro]

What happened

On PR #3746, the code agent modified public API exports in the boost plugin (added @backstage/plugin-permission-react dependency, new i18n keys, new utility exports) but did not regenerate report.api.md. This required two /fs-fix iterations (fix 1, fix 2) totaling approximately 25 minutes of agent compute time and two rounds of human intervention by mareklibra. The same pattern occurred on PR #3742 (ai-integrations workspace, 2 fix iterations) and PR #3743 (boost workspace). The boost workspace AGENTS.md lists yarn tsc:full && yarn build:api-reports:only as a build command, but this listing is descriptive rather than prescriptive — the code agent treats it as reference material, not as a mandatory verification step.

What could go better

The code agent should regenerate API reports as part of its verification phase whenever it modifies public exports. The existing workspace-specific AGENTS.md files list the command but don't frame it as mandatory. Issue #3757 proposes adding the instruction to the ai-integrations workspace AGENTS.md, but the problem recurs across multiple workspaces (boost, ai-integrations), making a repo-wide instruction more effective. Confidence: high — this is a clear, recurring pattern with three data points across two workspaces.

Proposed change

Add the following instruction to .fullsend/AGENTS.md, after the existing changeset guidance:

When adding, removing, or renaming public exports in any plugin package
(changes to `index.ts`, re-exported types, or API surface files), run
`yarn tsc:full && yarn build:api-reports:only` from the workspace root
(`workspaces/<workspace>/`) and include the updated `report.api.md`
files in your commit. CI will fail if API reports are stale.

This complements the workspace-specific fix proposed in #3757 by providing a repo-wide default that covers all current and future workspaces without requiring per-workspace AGENTS.md updates.

Validation criteria

The next 3 agent-authored PRs that modify public API exports (in any workspace) should include updated report.api.md files in their initial commit, without requiring /fs-fix iterations for API report regeneration.

---

Generated by retro agent from #3746

[fullsend-ai-triage]

🤖 Finished Triage · ✅ Success · Started 2:13 PM UTC · Completed 2:16 PM UTC

Commit: 5192aef · View workflow run →

[fullsend-ai-triage]

Relationship to #3757

This issue complements #3757, which proposes the same guidance at the workspace level (workspaces/ai-integrations/AGENTS.md). The two are not duplicates — this issue targets the repo-wide .fullsend/AGENTS.md to cover all current and future workspaces, while #3757 targets the ai-integrations workspace specifically. Both can proceed independently.

Proposed Test Case

Scenario: .fullsend/AGENTS.md contains API report regeneration instruction

Verify:
1. .fullsend/AGENTS.md contains instruction about running
   'yarn tsc:full && yarn build:api-reports:only'
2. Instruction specifies trigger: adding, removing, or renaming public exports
3. Instruction specifies working directory: workspaces/<workspace>/
4. Instruction mentions committing updated report.api.md files
5. Instruction warns CI will fail if API reports are stale

Edge cases:
- Instruction must be workspace-agnostic (not specific to boost or ai-integrations)
- Must not conflict with workspace-level AGENTS.md guidance from #3757

This is a small, well-scoped config edit with clear validation criteria — ready for implementation.

---

Labels: This is a feature/enhancement adding agent config instructions. 'fullsend' applies because the change targets .fullsend/AGENTS.md. 'Review effort 1/5' applies since it's a small config file edit. Removing 'ready-for-triage' and adding 'triaged' to reflect triage completion.

---

Next steps:

• /fs-code — agent creates a PR to implement this issue
• /fs-code <your instruction> — agent implements with your specific guidance