@gobing-ai/spur 0.4.9 → 0.4.11

package/config/workflows/task-pipeline.yaml

@@ -58,8 +58,8 @@ name: task-pipeline
 # verifier's and reviewer's self-graded levels (re-imported from the cc anti-hallucination skill
 # by task 1068) are no longer written, checked, or gated on; spur certifies on deterministic
 # evidence (rows MET, proof digests, findings dispositioned). The `confidence` check row, the
-# `ackLowConfidence` var, and the guards' LOW/absent clauses are gone; the derived
-# feature-receipt confidence (ADR-131/1196) is a different family and stays.
+# `ackLowConfidence` var, and the guards' LOW/absent clauses are gone. The derived
+# feature-receipt confidence (1196) followed in task 1221 (ADR-138 amendment).
 version: "7"
 description: "Standard task pipeline: precheck → implement → quality-gate (test/fix/recheck) → review → approve → verify → record → done"
 # Task entity lifecycle (owned by task-lifecycle.yaml):

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@gobing-ai/spur",
-    "version": "0.4.9",
+    "version": "0.4.11",
     "description": "Spur CLI — local-first harness for mainstream coding agents: constraint checking, workflow orchestration, agent health, and history analytics. Bun-native; exposes the `spur` command.",
     "keywords": [
         "spur",

package/plugins/sp/agents/expert-spur.md

@@ -98,7 +98,6 @@ their runbook here.
 **Noun(s):** task | feature | rule | workflow | agent spec
 **Method:** corpus campaign | compose/tune (`sp:spur-composer`) | evaluate/reflect (`sp:spur-doctor`)
 **Scope:** <resolved ids/files>
-**Confidence:** HIGH | MEDIUM | LOW
 
 ### Changes
 | Target | Operation | Result |

package/plugins/sp/commands/dev-changelog.md

@@ -26,5 +26,5 @@ For shared semantics, see the [flag glossary](../skills/spur-dev/references/flag
 
 ## Implementation
 
-Follow the inline procedure in [dev-operations.md](../skills/spur-dev/references/dev-operations.md#8-changelog) (changelog). This is a docs-only generator: run no quality gates (`bun run gate`, lint, typecheck, tests, `spur-check`, rules) — a generic "verify before finishing" rule does not apply here. The generator's own claim-verification pass is part of the procedure, not a gate, and carries two required checks per claim bullet (dev-operations §8, step 7): a **source citation** — an anchor naming a URL or a repo `path:line` — and a **confidence level**, exactly one of `HIGH`, `MEDIUM`, `LOW`, both held in the bullet's `<!-- src: … · <level> -->` comment and re-verified by step 8's self-check.
+Follow the inline procedure in [dev-operations.md](../skills/spur-dev/references/dev-operations.md#8-changelog) (changelog). This is a docs-only generator: run no quality gates (`bun run gate`, lint, typecheck, tests, `spur-check`, rules) — a generic "verify before finishing" rule does not apply here. The generator's own claim-verification pass is part of the procedure, not a gate, and carries one required check per claim bullet (dev-operations §8, step 7): a **source citation** — an anchor naming a URL or a repo `path:line` — held in the bullet's `<!-- src: … -->` comment and re-verified by step 8's self-check. No confidence level is recorded (ADR-138).
 

package/plugins/sp/commands/dev-pr-review.md

@@ -1,39 +1,49 @@
 ---
-description: "Review GitHub PRs with Codex; collect and fix findings."
+description: "Local-first PR review — status, a local branch-diff review, and validated fixes; --codex opts into the GitHub Codex review spine."
 role: reviewer
-argument-hint: "[full|submit|collect|fix|rerun|status|rules] [--base <branch>] [--no-wait] [--agent <inline|auto|name>] [<focus>]"
+argument-hint: "[status|review|fix] [--codex] [--force] [--no-wait] [--base <branch>] [--agent <inline|auto|name>] [<focus>]"
 allowed-tools: ["Bash", "Read", "Skill"]
 ---
 
 # Dev PR Review
 
-Wraps the **sp:pr-reviewing** skill. The review itself runs on the GitHub PR through Codex
-(`@codex review`) — never through a local Codex review mechanism; the invoking agent orchestrates
-and (in `fix` mode) implements.
+Wraps the **sp:pr-reviewing** skill. The default `status` mode is a read-only snapshot of the
+branch, its base and the pull request; `review` runs a **local** review of the branch diff unless
+`--codex` opts into the GitHub `@codex review` spine; `fix` triages findings from either reviewer
+and commits the verified ones. The invoking agent orchestrates.
 
 ## Argument Flags
 
 | Flag | Description | Default |
 | --- | --- | --- |
-| `<mode>` | `full\|submit\|collect\|fix\|rerun\|status\|rules` — full: request (deduped) + wait + report; submit: request and stop at pending; collect: report the latest review; fix: validate findings, fix legitimate ones, re-review; rerun: force a fresh review of the pushed HEAD; status: read-only composite; rules: author the repo's `AGENTS.md` `## Code Review Rules`. | full |
-| `--base` `<branch>` | Base branch when a new PR must be created. | existing PR base, else repo default |
-| `--no-wait` | Return pending right after the review request instead of polling. | off |
-| `--agent` `<inline\|auto\|name>` | Who performs model-bearing triage/fix. Omit uses the current agent, `inline` forbids dispatch, `auto` resolves the declared role, and a name pins that executor. This does not change the deterministic workflow/direct route. | omit |
-| `<focus>` | Remaining free text — extra review focus appended to the Codex request (e.g. `security boundaries and transaction idempotency`). | none |
+| `<mode>` | `status\|review\|fix` — status: read-only branch + PR state; review: local branch-diff review (or the GitHub Codex spine with `--codex`); fix: validate, fix and commit findings from either reviewer. | status |
+| `--codex` | `review` only — route through the GitHub Codex spine (`pr-review.yaml`) instead of the local `/sp:dev-review`. | off |
+| `--force` | `review --codex` only — request a fresh review of the current HEAD even when it was already reviewed (`mode=rerun`). | off |
+| `--no-wait` | `review --codex` only — post the request and return `pending` without polling (`mode=submit`). | off |
+| `--base <branch>` | Base branch for the branch report, the `pr-check` diff and a newly created PR. | existing PR base, else repo default |
+| `--agent <inline\|auto\|name>` | Who performs the model-bearing triage/fix. Omit uses the current agent, `inline` forbids dispatch, `auto` resolves the declared role, and a name pins that executor. This does not change the deterministic route. | omit |
+| `<focus>` | Remaining free text — extra review focus for the local review or the Codex request (e.g. `security boundaries and transaction idempotency`). | none |
 
 For shared semantics, see the [flag glossary](../skills/spur-dev/references/flag-glossary.md).
 
 ## Usage
 
-/sp:dev-pr-review [full|submit|collect|fix|rerun|status|rules] [--base <branch>] [--no-wait] [--agent <inline|auto|name>] [review focus]
+```
+/sp:dev-pr-review [status|review|fix] [--codex] [--force] [--no-wait] [--base <branch>] [--agent <inline|auto|name>] [review focus]
+```
 
 ## Implementation
 
 - Apply the [inline-default execution-surface contract](../skills/spur-dev/references/cross-cutting.md#inline-default-execution-surface).
 - Delegate everything: `Skill(skill="sp:pr-reviewing", args="$ARGUMENTS")`. The skill owns mode
-  routing, finding triage, fix, and rules authoring; `pr-review.yaml` is the SSOT
-  for the review spine's state order and guards; `spur builder pr` is the
-  deterministic git/gh core every spine step shells out to.
-- The external review goes through the GitHub PR and an `@codex review` comment only — never local
-  Codex review commands. Never force-push, rewrite history, merge the PR, or discard unrelated
-  changes; outside `fix` mode, ask before creating any commit.
+  routing, the local `/sp:dev-review` dispatch, the Codex spine and finding triage/fix. On the
+  Codex path `pr-review.yaml` is the SSOT for state order and guards, and the four flat
+  `spur builder pr-<verb>` verbs are the deterministic git/gh core every spine step shells out to.
+- Retired modes print one notice line and run their replacement: `full` → `review --codex`;
+  `submit` → `review --codex --no-wait`; `rerun` → `review --codex --force`; `collect` → `status`.
+  `rules` prints that `AGENTS.md` review-rule authoring belongs to `cc:cc-magents` (the `cc` plugin)
+  and exits without running a review.
+- Never force-push, rewrite history, close or integrate the PR, or discard unrelated changes.
+  `review` without `--codex` makes no GitHub writes. `fix` pre-authorizes one focused commit for the
+  verified findings — stop if unrelated uncommitted changes make safe separation or committing
+  ambiguous.

package/plugins/sp/plugin.json

@@ -1,5 +1,5 @@
 {
     "name": "sp",
-    "version": "0.4.9",
+    "version": "0.4.11",
     "description": "Spur — a local-first harness engineering toolkit that wraps mainstream coding agents with constraint checking, workflow orchestration, and history analytics."
 }

package/plugins/sp/skills/brainstorm/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: brainstorm
-description: "Structured ideation: generate solution options with trade-offs and confidence scoring, then delegate to research/task-creation skills. Triggers: \"brainstorm ideas\", \"explore solutions\", \"consider options\", \"research approaches\", \"what are my options\", \"how should I approach X\"."
+description: "Structured ideation: generate solution options with trade-offs and source citations, then delegate to research/task-creation skills. Triggers: \"brainstorm ideas\", \"explore solutions\", \"consider options\", \"research approaches\", \"what are my options\", \"how should I approach X\"."
 license: Apache-2.0
 version: 1.0.0
 created_at: 2026-03-25
@@ -30,7 +30,7 @@ see_also:
 
 # sp:brainstorm — Structured Ideation Workflow
 
-Generate solution options with trade-offs, recommendations, and confidence scoring. Delegates research to specialized skills.
+Generate solution options with trade-offs, recommendations, and source citations. Delegates research to specialized skills.
 
 **Key distinction:**
 - **`sp:brainstorm`** = Ideation: generate approaches with trade-offs
@@ -40,7 +40,7 @@ Generate solution options with trade-offs, recommendations, and confidence scori
 
 ## Overview
 
-The `sp:brainstorm` skill generates multiple solution approaches with explicit trade-offs, confidence scoring, and source citations. It follows a structured 3-phase workflow: Input parsing, Ideation with research delegation, and structured Output. Unlike pure research or bare task creation, brainstorm focuses on ideation—generating and comparing options before committing to a solution path.
+The `sp:brainstorm` skill generates multiple solution approaches with explicit trade-offs and source citations. It follows a structured 3-phase workflow: Input parsing, Ideation with research delegation, and structured Output. Unlike pure research or bare task creation, brainstorm focuses on ideation—generating and comparing options before committing to a solution path.
 
 ## Quick Start
 
@@ -101,14 +101,15 @@ See the [inline-default contract](../spur-dev/references/cross-cutting.md#inline
 
 ### 3. Generate 2-3 Approaches
 
-Always generate multiple options, each with description, trade-offs, confidence, and sources — full
+Always generate multiple options, each with description, trade-offs, and sources — full
 per-approach template: [references/workflows.md](references/workflows.md#approach-generation).
 
-### 4. Confidence Scoring
+### 4. Source Citations
 
-Every approach and every external claim carries a HIGH/MEDIUM/LOW confidence score plus a dated
-source citation — table, thresholds, and citation format:
-[references/workflows.md](references/workflows.md#confidence-scoring).
+Every external claim carries a dated source citation — format:
+[references/workflows.md](references/workflows.md#source-citation-format). No HIGH/MEDIUM/LOW
+confidence score (ADR-138): a self-assigned score is as easy to write as the claim; the citation and
+the trade-offs are what the operator compares.
 
 ### 5. Task Delegation
 
@@ -224,13 +225,13 @@ still pauses unless the operator has encoded prior approval in the workflow vars
 | "I understand the request — skip clarifying." | Ideating on a misread wastes the whole session. Validate input clarity and the two input modes before generating options. |
 | "More options are always better." | Beyond ~3, options dilute focus and stall the decision. Cap at three and delegate deeper research instead. |
 | "I'll design the architecture while I brainstorm." | Brainstorm generates and scores approaches; it does not lock the design. Route a chosen approach through the Design Approval Gate. |
-| "Confidence scores are subjective — skip them." | An unscored option hides its risk. Cite sources and assign confidence so the operator compares on evidence, not vibes. |
+| "Sources are obvious — skip them." | An uncited option hides its risk. Cite sources and name the trade-offs so the operator compares on evidence, not vibes. |
 
 ## Red Flags
 
 - Presenting a single approach with no alternatives or trade-offs.
 - Starting ideation before the input is clarified.
-- More than three approaches, or approaches with no confidence score / source.
+- More than three approaches, or approaches with no source.
 - Skipping the task-delegation offer after the operator confirms a direction.
 - Treating a brainstorm output as an approved design (bypassing the approval gate).
 
@@ -293,7 +294,7 @@ codebase-first rule, stop conditions, depth question budgets, smart positional d
 The skill uses this tree to:
 1. **Constrain the option space** — each approach must be compatible with resolved decisions
 2. **Generate decision-trace annotations** — each approach lists which decisions it depends on
-3. **Calibrate confidence** — decisions resolved from codebase evidence get higher confidence
+3. **Cite the evidence** — decisions resolved from codebase evidence cite the `path:line` they rest on
 
 When `--skip-discovery` is used (no `--context`), the operation falls back to the standard
 3-phase workflow with its own lightweight clarification step.

package/plugins/sp/skills/brainstorm/agents/openai.yaml

@@ -1,6 +1,6 @@
 name: brainstorm
 description: "Structured ideation workflow for generating solution options with
-  trade-offs, confidence scoring, and delegation to research and task creation
+  trade-offs, source citations, and delegation to research and task creation
   skills. Triggers: brainstorm ideas, explore solutions, consider options,
   research approaches, multiple solution options with trade-offs."
 version: 1.0.0

package/plugins/sp/skills/brainstorm/examples/ideation-example.md

@@ -80,7 +80,6 @@ async function authMiddleware(c: Context, next: Next) {
 }
 ```
 
-**Confidence:** HIGH
 **Sources:**
 - [jose Library Docs](https://github.com/panva/jose) | Verified: 2026-03-25
 - [Hono Middleware Guide](https://hono.dev/docs/middleware/builtin/bearer-auth) | Verified: 2026-03-25
@@ -127,7 +126,6 @@ async function handleLogin(username: string, password: string) {
 }
 ```
 
-**Confidence:** MEDIUM
 **Sources:**
 - [Auth0 Node.js SDK v5](https://auth0.github.io/node-auth0/) | Verified: 2026-03-25
 - [Firebase Admin Setup](https://firebase.google.com/docs/admin/setup) | Verified: 2026-03-15
@@ -172,7 +170,6 @@ async function apiKeyMiddleware(c: Context, next: Next) {
 }
 ```
 
-**Confidence:** MEDIUM
 **Sources:**
 - [Hono Bearer Auth Middleware](https://hono.dev/docs/middleware/builtin/bearer-auth) | Verified: 2026-03-25
 
@@ -206,4 +203,3 @@ async function apiKeyMiddleware(c: Context, next: Next) {
 
 **Generated by:** sp:brainstorm
 **Research delegation:** sp:source-driven-development, spur agent run
-**Confidence:** HIGH (official jose/Hono documentation, 2026)

package/plugins/sp/skills/brainstorm/metadata.openclaw

@@ -1,6 +1,6 @@
 name: brainstorm
 description: "Structured ideation workflow for generating solution options with
-  trade-offs, confidence scoring, and delegation to research and task creation
+  trade-offs, source citations, and delegation to research and task creation
   skills. Triggers: brainstorm ideas, explore solutions, consider options,
   research approaches, multiple solution options with trade-offs."
 version: 1.0.0

package/plugins/sp/skills/brainstorm/references/workflows.md

@@ -85,7 +85,7 @@ DO NOT implement research directly. Delegate to specialized skills:
 
 2. Delegate research + synthesis inline by default
    - Run in the current session; escalate to `spur agent run` only on a subprocess trigger
-   - Request confidence scoring
+   - Request source citations
 ```
 
 ### Approach Generation
@@ -108,7 +108,6 @@ DO NOT implement research directly. Delegate to specialized skills:
 - Key technical considerations
 - Dependencies or prerequisites
 
-**Confidence:** HIGH/MEDIUM/LOW
 **Sources:** [Citations with dates]
 ```
 
@@ -120,14 +119,6 @@ DO NOT implement research directly. Delegate to specialized skills:
 - [Title](URL) | **Verified**: YYYY-MM-DD
 ```
 
-### Confidence Scoring
-
-| Level | Range | Criteria |
-|-------|-------|----------|
-| **HIGH** | >90% | Direct quote from official docs (2025+) |
-| **MEDIUM** | 70-90% | Synthesized from multiple sources |
-| **LOW** | <70% | Uncertain, flag for review |
-
 ---
 
 ## Phase 3: Structured Output
@@ -144,7 +135,7 @@ DO NOT implement research directly. Delegate to specialized skills:
 2. **Approaches** (200-300 words each)
    - 2-3 options with trade-offs
    - Implementation considerations
-   - Confidence levels
+   - Sources
 
 3. **Recommendations** (100-150 words)
    - Recommended approach with reasoning
@@ -197,7 +188,6 @@ run_id: <run id, omit when none>
 **Implementation Notes:**
 - [Technical considerations]
 
-**Confidence:** HIGH/MEDIUM/LOW
 **Sources:** [Citations]
 
 ### Approach 2: [Name]
@@ -255,7 +245,7 @@ Bash: spur task batch-create --file decomposition.json   # bare JSON array (see
 | File not found | Provide clear error with path |
 | Invalid YAML | Report parsing error |
 | Missing fields | List missing fields, ask if incomplete |
-| Research unavailable | Continue, note reduced confidence |
+| Research unavailable | Continue, mark the uncited claims |
 | Save fails | Display output, suggest manual save |
 
 ---
@@ -265,6 +255,6 @@ Bash: spur task batch-create --file decomposition.json   # bare JSON array (see
 When tools unavailable:
 
 1. **Continue** with available tools
-2. **Note** reduced confidence
+2. **Mark** claims left uncited
 3. **Suggest** manual fallback
 4. **Save** intermediate state

package/plugins/sp/skills/code-verification/references/secu-review.md

@@ -90,10 +90,13 @@ Before declaring a task `done`, run this lightweight checklist. It catches the m
 - [ ] No `console.log` / `console.error` in production code (use the project logger).
 - [ ] No `--no-verify`, `--force`, or new suppression comments added to bypass gates.
 - [ ] Solution section contains `file:line` citations for every changed file.
-- [ ] API/library claims in changed docs carry a source citation (URL or `path:line`) and an honest
-      confidence level (HIGH/MEDIUM/LOW) per source-driven-development — a HIGH claim must show the
-      source it was verified against on the same line (MEDIUM/LOW need no anchor); the deterministic advisory backstop is
-      the `cited-version-strings` / `cited-api-claims` / `cited-api-claims-confidence` / `cited-high-confidence` rules
+- [ ] Every API/library claim in changed docs carries BOTH a source citation (URL or `path:line`) and
+      an EARNED level `HIGH|MEDIUM|LOW` — `HIGH` cites (a URL, `path:line`, or a path with a
+      code/doc extension), `MEDIUM`/`LOW` cite or name what is unverified. That is the
+      `| Claim | Source | Local verification | Confidence |` contract of
+      `.spur/rules/structure/api-library-claim-citation.yaml`; ADR-138 still forbids a confidence
+      label on the VERDICT or verify answer. The deterministic advisory backstop for the citation
+      half is the `cited-version-strings` / `cited-api-claims` rules
       (`.spur/rules/docs/sourced-api-claims.yaml`, run ad hoc — not wired into a preset).
 - [ ] Review section has P1–P4 findings table (even if all rows are empty — the table itself proves review happened).
 

package/plugins/sp/skills/history-anatomy/references/operations.md

@@ -31,10 +31,9 @@ Given the current and baseline artifacts, author the model half of the report. A
    calendar day; ad-hoc → immediately preceding equal-duration window. If baseline coverage is
    insufficient or materially different, emit `not comparable` and **no** trend/delta/percentage.
 3. **Findings** — each with the full per-finding field set (key, category, impact, trend,
-   observation, inference, confidence, contradictions, evidenceAnchor, severity, reproCommand,
+   observation, inference, contradictions, evidenceAnchor, severity, reproCommand,
    ownerSurface). Categories from the closed vocabulary; stable keys of the form
-   `<category>:<owner-surface>:<signal>`. Severity is the closed P1/P2/P3 vocabulary and is
-   orthogonal to confidence; reproCommand reproduces the observation verbatim; ownerSurface names
+   `<category>:<owner-surface>:<signal>`. Severity is the closed P1/P2/P3 vocabulary; reproCommand reproduces the observation verbatim; ownerSurface names
    the concrete owning surface (file/package/command), consistent with the key's middle segment.
 4. **Recurrence ledger** — classify every finding against the baseline on the **stable key**.
 5. **Telemetry gaps** — every dimension the artifact cannot support, rendered `not available`.
@@ -58,7 +57,7 @@ Given a candidate report and the artifacts, independently verify:
 - **Twelve-section completeness and order** — all twelve section names present in the frozen
   order; none renamed/omitted.
 - **Per-finding fields** — every finding (problem and positive) carries key, category, impact,
-  trend, observation, inference, confidence, contradictions, evidenceAnchor, severity (P1/P2/P3),
+  trend, observation, inference, contradictions, evidenceAnchor, severity (P1/P2/P3),
   reproCommand, ownerSurface; category in the closed vocabulary; stable-key grammar. A finding
   missing any triage field FAILs (the deterministic gate enforces the same set).
 - **Remediation handoff** — each accepted-proposal handoff carries a `spur task create`

package/plugins/sp/skills/history-anatomy/references/report-contract.md

@@ -62,10 +62,9 @@ Every finding — problem and positive alike — carries the full field set:
 | `trend` | `new` / `recurring` / `regressed` / `improved` / `resolved` / `not-comparable` (see recurrence). |
 | `observation` | What the artifacts show — evidence, not interpretation. |
 | `inference` | What the observation is reasoned to mean; names its supporting observations. |
-| `confidence` | Per-finding: `high` / `medium` / `low`. Never one blanket report-level score. |
 | `contradictions` | Any contradicting signal shown beside the finding, not silently reconciled. |
 | `evidenceAnchor` | At least one anchor to the forensics artifact or cited `file:line`. An entry with no anchor is invalid. |
-| `severity` | Closed vocabulary `P1` / `P2` / `P3` — how much the finding matters. Orthogonal to `confidence`: severity orders work, confidence says how sure we are (0680 R1). The structure gate fails a finding without one. |
+| `severity` | Closed vocabulary `P1` / `P2` / `P3` — how much the finding matters. Severity orders work (0680 R1); how well a finding is supported lives in its `observation`, `inference`, `contradictions` and `evidenceAnchor`, never a confidence label (ADR-138). The structure gate fails a finding without one. |
 | `reproCommand` | The invocation that reproduces the observation — one command a reader can run verbatim (0680 R2). Gate-enforced. |
 | `ownerSurface` | The concrete surface that owns the fix (a file path, package, or command), consistent with the key's `<owner-surface>` segment but named as a target rather than a slug (0680 R3). Gate-enforced. |
 
@@ -114,7 +113,7 @@ reclassify a recurring finding as new.
 ## Positive patterns
 
 Positive entries are held to the **same evidence standard as problems**: they carry observation,
-inference, confidence, and at least one evidence anchor. An entry without an anchor is invalid. The
+inference, and at least one evidence anchor. An entry without an anchor is invalid. The
 section renders successful workflows, resolved past issues that stay resolved, and healthy,
 repeating behavior worth keeping.
 

package/plugins/sp/skills/issue-finding/SKILL.md

@@ -411,7 +411,7 @@ Thin wrapper: `Skill(skill="sp:issue-finding", args="$ARGUMENTS")`.
 - Wholesale raw JSONL parsing when the source has a typed mapper.
 - Accepting `--use-history` or `--no-task` silently instead of naming their replacements.
 - GENERATE recipes inventing a title flag, dotted template forms, or quoted dotted section flags.
-- Claiming multi-agent forensics without stating source confidence (High/Medium/Low).
+- Claiming multi-agent forensics without citing the session source each finding came from.
 - Skipping batch section writes + single `spur task check` (when generating).
 - Emitting empty findings without inventorying sessions first.
 

package/plugins/sp/skills/parallel-execution/SKILL.md

@@ -61,7 +61,7 @@ Parallel subagent outputs must be **synthesized**, not concatenated. The synthes
 
 1. **Dedup** — merge findings with the same `file:line` anchor; keep the highest-severity version.
 2. **Resolve conflicts** — when two subagents disagree on the same claim, surface the disagreement explicitly; don't silently pick one.
-3. **Rank by confidence** — sort synthesized results by confidence (HIGH → MEDIUM → LOW); unresolved conflicts go last.
+3. **Rank by evidence** — verified findings (executed check or cited source) first, unverified next, unresolved conflicts last; severity within each tier.
 4. **Unified format** — emit one coherent output, not N raw dumps.
 
 Full synthesis methodology: [result-synthesis.md](references/result-synthesis.md).

package/plugins/sp/skills/parallel-execution/references/result-synthesis.md

@@ -41,13 +41,12 @@ Synthesis:  ⚠ CONFLICT — Redis migration safety
 
 **Rule:** conflicts are findings, not failures. Surface them with both positions stated; let the operator resolve.
 
-### 3. Rank by confidence
+### 3. Rank by evidence
 
-Sort synthesized results:
-1. HIGH confidence (empirically verified, source-cited)
-2. MEDIUM confidence (reasoned, but not empirically verified)
-3. LOW confidence (heuristic, pattern-match, or uncertain)
-4. Unresolved conflicts (both sides stated)
+Sort synthesized results by what backs them, not a self-assigned confidence label (ADR-138):
+1. Verified (an executed check or a cited `path:line` / URL)
+2. Unverified (reasoned or pattern-matched; name the check that would ground it)
+3. Unresolved conflicts (both sides stated)
 
 Within each tier, sort by severity (P1 → P2 → P3 → P4).
 
@@ -58,11 +57,11 @@ Emit one coherent output. The unified report has:
 ```
 ## Synthesized Results
 
-### HIGH Confidence
-| # | Finding | Source | Severity | File:Line |
-|---|---------|--------|----------|-----------|
+### Verified
+| # | Finding | Source | Severity | File:Line | Evidence |
+|---|---------|--------|----------|-----------|----------|
 
-### MEDIUM Confidence
+### Unverified
 ...
 
 ### Conflicts (unresolved)