The user wants more leads on the current lens — a bigger batch than what the standard daily fill delivers. Execute the extend flow end-to-end.

GATE — DEFER TO TOOL RENDERING. When you call a Leadbay composite that ships its own RENDERING block (every composite in 0.9.0+ does), render the response using that block's recipe verbatim — score bars, glyph palette, column order, hide-list, link priorities, all of it. Do NOT substitute prose, a numbered list, or a different column structure even when an orchestrating prompt's body suggests alternate framing. Prompt-specific commentary (motivational nudges, summaries, next-action recommendations) belongs ABOVE or BELOW the canonical table, never in place of it.

If the prompt's body and the tool's RENDERING appear to conflict, the tool's RENDERING wins for the structural layout; the prompt's voice wins for the commentary that surrounds it.

PHASE 1 — PRE-CHECK QUOTA

Call leadbay_account_status and read quota.org.resources[] for the LENS_EXTRA_REFILL entry. If count is at/over the daily cap and extra_count (when provided) wouldn't fit, skip PHASE 2 and surface the three options via your host's choice widget (`ask_user_input_v0` or `AskUserQuestion`) now: smaller count / wait until resets_at / upgrade plan. Don't waste a write call you already know will 429.

PHASE 2 — PICK SEEDS SILENTLY

IRON LAW — DO NOT SHOW THE SEED CANDIDATES TO THE USER. They asked for more leads, not a candidate review. The seed list is internal scaffolding.

Call leadbay_seed_candidates (defaults to last-active lens). Pick 3–5 seeds using this priority order:

1. Engagement (load-bearing) — liked: true, then high org_contacts_count + prospecting_actions_count. The user already validated these.
2. `qq_answers` — pick candidates whose qualification answers align with the target profile.
3. `tags` — purchase-intent alignment.
4. `sector / size_min / size_max` — shape similarity.
5. `ai_agent_score` — overall AI fit (tie-breaker).

Default heuristic when nothing else differentiates: top 3 by engagement (prefer liked; break ties by combined action count), then fill to 5 with the top ai_agent_score rows whose tags overlap the engagement leaders.

PHASE 3 — FIRE THE EXTEND

Call leadbay_extend_lens with:

• seed_lead_ids: the picked lead_ids from PHASE 2
• extra_count: <How many extra leads to add. Optional. Omit to use the backend default. Optional.> when set, otherwise omit (backend default)
• lensId: omit (uses the same default as PHASE 2)

PHASE 4 — REACT TO STATUS

Never list accepted_seeds to the user. They are internal — the user wants the outcome (queue confirmation), not the picking step.