6. Scaffold from the brief¶

Step 6 of 10 · Deliver to a customer

Step at a glance

🎯 Goal — Translate the brief into code, infra, evals, telemetry.

📋 Prerequisite — 5. Discover with the customer complete — docs/discovery/solution-brief.md has zero TBD.

💻 Where you'll work — VS Code (Copilot Chat sidebar + Source Control panel for diff review).

✅ Done when — /scaffold-from-brief has run; the diff is reviewed and committed; python scripts/accelerator-lint.py passes.

Chatmodes used here

/scaffold-from-brief

Full reference: Chatmodes overview.

What success looks like

git status after the scaffold run shows changes spread across (typical):

modified:   src/scenarios/sales_research/agents/supervisor/prompt.py
modified:   src/scenarios/sales_research/retrieval.py
new file:   src/tools/<your-new-tool>.py
modified:   accelerator.yaml
modified:   evals/quality/golden_cases.jsonl
modified:   evals/redteam/<scenario>.jsonl
modified:   infra/main.parameters.json

python scripts/accelerator-lint.py finishes with:

✅ accelerator-lint: 30/30 rules passed

In Copilot Chat:

/scaffold-from-brief

Copilot reads the filled brief and customises the repo. The Lands in column below shows paths for the flagship scenario (sales-research).

If you scaffold a new scenario via python scripts/scaffold-scenario.py <scenario-id> (e.g., customer-service, rfp-response), substitute <scenario-id> for sales_research in the src/scenarios/<...>/ paths. Everything outside src/scenarios/ is scenario-agnostic and stays put.

Brief field →	Lands in (flagship paths shown; `src/scenarios/<id>/` for custom scenarios)
Problem + persona	`src/scenarios/sales_research/agents/supervisor/prompt.py` system prompt
Solution shape	Keep flagship OR run `/switch-to-variant` for a walkthrough of re-authoring under `patterns/single-agent` or `patterns/chat-with-actioning` (manual re-authoring walkthroughs, not drop-ins)
Grounding sources	`src/retrieval/ai_search.py` (scenario-agnostic client) + scenario-specific index schema at `src/scenarios/sales_research/retrieval.py` + `infra/modules/ai-search.bicep`
Side-effect tools	New files under `src/tools/` with HITL scaffolding
HITL gates	Per-tool `HITL_POLICY` constant + `checkpoint(...)` calls; `accelerator.yaml -> solution.hitl` engagement-level summary
Constraints	`infra/main.parameters.json` + `accelerator.yaml`
Success criteria	`evals/quality/golden_cases.jsonl` + CI gates
RAI risks	`evals/redteam/` custom adversarial cases
ROI KPIs	`src/accelerator_baseline/telemetry.py` events + `infra/dashboards/roi-kpis.json` (panels are scenario-agnostic; rename the dashboard per engagement)

Re-run /scaffold-from-brief whenever the brief changes — the same expansion reapplies across every artefact.

Authoring agent instructions¶

Agent system instructions live in docs/agent-specs/<agent>.md under the ## Instructions heading — edit those Markdown files, not Python. On azd up (next step), the FastAPI startup bootstrap (src/bootstrap.py) syncs each spec verbatim to the Foundry portal once the Container App boots. prompt.py is for per-request input construction only.

Review the diff¶

Open VS Code's Source Control panel (Ctrl+Shift+G). Walk every changed file. Then:

git add -A
git commit -m "Scaffold from brief"
python scripts/accelerator-lint.py

The lint runs 30 deterministic rules against accelerator.yaml + repo state — manifest shape, per-agent model references, content-filter attachment, HITL coverage on side-effect tools, workflow secrets documented, deploy chain gating, and more. CI re-runs it on every PR, but running it locally first saves a round-trip.

If the lint flags anything, fix it now — every later step assumes the lint is green.

Continue → 7. Provision the customer's Azure