[pm25coder]

Closes #2581

What this changes

emrg/server/promote_prompt.md, two places, one rule:

• The summary template's blocked field (§4.1) now states one availability rule. It kept the per-channel requirement it was given and lost the clause empty = no blocker, which that requirement makes unreachable: an account holding one channel owes that channel a line in every round, so the field is empty only when there is no channel at all. What replaces it says the same thing the retired clause was trying to say — no blocker is stated per line, never by leaving the field empty.
• The zone list (§4.4, "C. Channel status") no longer restates the rule. It keeps its own job — naming blocked as the single field that owns channel availability — and points at the form §4.1 gives it. The template's §4.2 states why: a fact written twice will eventually disagree with itself.

The per-channel sentence itself is untouched, deliberately: tests/test_prompt_templates.py::PROMOTE_METHOD_RULES pins that phrase's position and its single occurrence, and a rewrite that reworded it would have moved a guard nobody asked to move.

The guard

Four new tests in tests/test_prompt_templates.py, in the section shape that file already uses:

• the field carries the per-channel rule;
• the template states no clause the field cannot satisfy, and says how "no blocker" is recorded instead;
• the phrase the rule is written in occurs once, in the field that owns it, and not in the zone list. Named limit: this counts that phrase, so a restatement in other words would pass.
• a discriminating test whose arms are built from the live text with each premise asserted first, so an arm whose edit landed nowhere fails loudly instead of measuring nothing.

Verification, on the tree this PR lands

Measured 2026-10-26 on base ad8f44c (master) with this branch merged in — the head is 4ecaf86 locally, 89f58d7d on the remote, content identical.

• tests/test_prompt_templates.py: 54 passed (50 before this change + 4 new).
• Full suite: 5385 passed, 512 skipped (908.92s, exit 0). The suite read 5381 before this change, so the delta is exactly the four new tests.
• scripts/check-doc-count.py: exit 0 — no tracked file states the Python test count; --measure reads 5893 collected tests at ad8f44c.
• python -m emrg --help: exit 0. from emrg.client.app import run_client: imports.
• Five mutation arms, each restored afterwards to a byte-identical file (git hash-object matched the pre-mutation blob every time):

Arm	Mutation	Guard that failed
A1	put empty = no blocker back into the field	test_the_blocked_field_states_no_clause_it_cannot_satisfy
A2	every channel → each channel	test_the_blocked_field_carries_the_per_channel_rule
A3	copy the rule back into the zone bullet	test_the_per_channel_blocked_rule_has_one_home
A4	drop the replacement clause	test_the_blocked_field_states_no_clause_it_cannot_satisfy (second assertion)
A5	remove "not run" from the field	the first and third guards, plus the arms test — the rule left its only home

CI on this head: pending as this PR is opened; it is not reported as green.

[pm25coder]