fix(advisor): the front door was a thin caller, not a structural dead end

[stranske]

Stacks on #17 (same file), so the base is #17's branch — the diff here is only this change. Merge #17 first.

The claim I got wrong

I reported that 35 of 41 capabilities were structurally unreachable from the advisor. That was wrong, and it mattered — it framed the central problem of this whole effort as a design limit to work around instead of a caller to fix.

capabilities._matches_trigger matches a {"kind": k, "name": n} matcher against a same-named field the caller supplies, and its own comment says so: "Adding a new trigger kind is then a caller-side change, not an edit here." The advisor was handing it {repository, task_type, lane} and nothing else.

adversarial-review, matcher {kind: closer_gate, name: high_stakes_review}
  advisor trigger today   -> (False, ['closer_gate_not_in_trigger'])
  same matcher, one field -> (True, [])

What changed

1. One constant. DIRECT_ENTRY was {"offload": "offload"} under a comment claiming it mirrored dispatcher.TASK_TYPE_CAPABILITY. It didn't — the dispatcher routed runtime_ac to runtime-ac-checks while the advisor named deliberate-break-verifier for the same work. Now derived from the dispatcher.
2. Context forwarded (context=, --context), confined to kinds the ledger declares. Absent context still fails closed.
3. The whole denominator. Every non-match reports the reason _matches_trigger already returned, plus an entry_requirement derived from its own matcher. Those reasons were computed and discarded — that discard was the silence.

Numbers

	before	after
reachable, free text	6	7
reachable, real caller context	6	12
accounted for with a reason	6 of 41	41 of 41

No matcher rewritten, no task-signal widened. Widening keywords to move the number would corrupt the associations this feeds.

Break→revert

• Reverting DIRECT_ENTRY: fails on dispatcher routes 'runtime_ac' to 'runtime-ac-checks' but the advisor does not name it.
• Removing reason-reporting: fails with all 39 unaccounted capabilities named.

verify.py: 353 passed, 0 failed, 0 skipped, 81/81 selftests, 5/5 gates. The repo's own test_verifying_the_system_never_writes_the_live_ledger guard caught a writing load() in my selftest — fixed to load_declared().

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro

Run ID: ef33cd61-b6c5-4229-8da9-7d518ee75f39

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}