fix(discovery): give the research Effort-ceiling sentence headroom

.changes/discovery/work-resolve-issue-issue-12631-41571316834-1-f83c6418.md

@@ -0,0 +1,7 @@
+---
+bump: patch
+---
+
+### Changed
+
+- `/discovery:research`: the sentence saying the Effort row caps discipline 8 now sits in the Disciplines intro, ahead of the numbered list, instead of opening "Effort, source breadth". It now ends at about byte 15,200, well inside the first 20,000 bytes that compaction re-attaches. The paragraph on grading each sub-slice of an N-topic fan-out moved under "The deep tier", and acceptance gate step 2 points to it there.

plugins/discovery/scripts/contract.test.sh

@@ -386,6 +386,12 @@ assert_in_slice 'research effort ceiling heading is inside the re-attach slice'
   'skills/research/SKILL.md' '### Effort, source breadth'
 assert_in_slice 'research effort ceiling sentence is inside the re-attach slice' \
   'skills/research/SKILL.md' 'The Effort row is the ceiling over discipline 8'
+# #12631 asks for 2,000 bytes of headroom, so the whole sentence ends by byte 18,000.
+if [[ "$(head -c 18000 "$PLUGIN_ROOT/skills/research/SKILL.md")" == *'The Effort row is the ceiling over discipline 8: "Effort, source breadth" below.'* ]]; then
+  pass 'research effort ceiling sentence keeps 2,000 bytes of headroom'
+else
+  fail 'research effort ceiling sentence keeps 2,000 bytes of headroom'
+fi
 assert_in_slice 'research topic slot is inside the re-attach slice' \
   'skills/research/SKILL.md' 'Research the following topic: $ARGUMENTS'
 if grep -q '^## Phase 0:' "$PLUGIN_ROOT/skills/research/context/phases.md" \

plugins/discovery/skills/research/SKILL.md

@@ -39,7 +39,7 @@ The disk fallback Reads this same file, so a matching token is file-identity, **
 
    Cite the **exit status**, 0 usable, 1 no usable artifact set, 2 ungradeable, not a reading of the directory: the context most motivated to call the dispatch finished would be the reader. Only the slice path and `--index-name` are required; the bare form is still a real gate, each optional check reporting `unchecked`, never a quiet pass. Append `--expect-sidecars <n>` when the payload reported a `sidecars:` count, and **drop any flag whose value the payload did not supply**. **The `index=` path in that output is authoritative** downstream: the verifier's `target` and the handoff pointer come from it, not from `artifact:`.
 
-   **Fanning out over N topics, grade each run against the sub-slice IT was assigned, before synthesizing the slice-root index.** The gate grades exactly the path it is given, so a slice-root invocation grades only the synthesis, never a dispatched run. The synthesis then goes to a fresh verifier for the verifier-owned rows before it is surfaced. A slice-root synthesis that adds no claims and cites by ID needs no verifier: run the restatement check with `--sub-slice` per sub-slice, check its `accepted:`, and if both pass write `verification: skipped (cost)`. Both: the dispatch contract's fan-out section.
+   On an N-topic fan-out, grade each sub-slice before the synthesis, then send the synthesis to a fresh verifier for the verifier-owned rows (the one exception, and the rest: "The deep tier" below).
 
 3. **The coverage claim is graded from the ledger, not from the payload.** `coverage: complete` mirrors outcome-gate criterion 11, which the run graded on **itself**. When a `research-checklist.md` sits beside the index step 2 named, run `"${CLAUDE_PLUGIN_ROOT}/scripts/check-coverage-complete.sh"` (or the `.py` twin) on `<that index's directory>/research-checklist.md` and cite its exit status. 0 complete, 1 unmarked rows, 2 ungradeable, and **both non-zero values are FAILs**. No ledger on disk is correct **only** when the artifact records the corpus as unbounded; a bounded corpus with no ledger is a Phase 0 that never ran, whatever the payload says.
 
@@ -100,7 +100,7 @@ Non-topic tokens (`breadth=`, `deep`, `--unattended`), a dispatched run, and a r
 
 ## Disciplines
 
-Full recipes and rationale: `${CLAUDE_PLUGIN_ROOT}/skills/research/context/discipline.md` (also the canonical source-tier table for this plugin).
+Full recipes and rationale: `${CLAUDE_PLUGIN_ROOT}/skills/research/context/discipline.md` (also the canonical source-tier table for this plugin). The Effort row is the ceiling over discipline 8: "Effort, source breadth" below.
 
 1. **3 phases minimum**. Phase 1 (broad), Phase 2 (targeted, informed by Phase 1, includes falsification), Phase 3 (preferred-sources / tool-ecosystem fallback)
 2. **Queries scale to open questions: the floor is a starting point, not a target.** Phase 1 opens with ≥3 queries to seed the evidence base; Phase 2 and Phase 3 each run **one query per unresolved gap/conflict** surfaced by the prior phase's written analysis (≥3, no upper cap). A run that stops at the floor while numbered gaps remain has not finished the phase
@@ -120,8 +120,7 @@ Full recipes and rationale: `${CLAUDE_PLUGIN_ROOT}/skills/research/context/disci
 
 ### Effort, source breadth
 
-The Effort row is the ceiling over discipline 8. Rationale and skipped-phase N/A: the discipline
-file's "Effort, source breadth".
+Rationale for the ceiling over discipline 8, and skipped-phase N/A: the discipline file's "Effort, source breadth".
 
 Caller effort for this run is `${CLAUDE_EFFORT}`. If that still begins with `${`, this body was read
 directly and the substitution never ran: run as `high`, every phase below. A value the table below
@@ -164,6 +163,8 @@ budget, and an ungradeable or missing artifact still takes the recovery ladder,
 In the main conversation, take the deep tier when any of these holds, and read [context/deep-tier.md](context/deep-tier.md) before dispatching anything:
 
 - **Two or more separable topics** (they share no claims): one `discovery:researcher` per topic, each in its own sub-slice, never one engine over the combined blob.
+
+  **Fanning out over N topics, grade each run against the sub-slice IT was assigned, before synthesizing the slice-root index.** The artifact gate (acceptance gate step 2) grades exactly the path it is given, so a slice-root invocation grades only the synthesis, never a dispatched run. The synthesis then goes to a fresh verifier for the verifier-owned rows before it is surfaced. A slice-root synthesis that adds no claims and cites by ID needs no verifier: run the restatement check with `--sub-slice` per sub-slice, check its `accepted:`, and if both pass write `verification: skipped (cost)`. Both: the dispatch contract's fan-out section.
 - **One topic that needs adversarial verification**: contested, high-stakes, or broad enough (multi-vendor, comparison, migration) that independent skeptics should try to refute each claim. The `discovery:research-sweep` workflow runs it when the Workflow tool is observed in this session's toolset; otherwise one researcher does.
 - **A leading `deep` token** in the arguments: strip it from the topic, then route as above. A `breadth=` token with it wins, and the run takes the standard dispatch.