HelloThisWorld · HelloThisWorld · Aug 30, 2026 · Aug 30, 2026 · Aug 30, 2026 · Aug 30, 2026
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -2,6 +2,28 @@
 
 ## Unreleased
 
+### Telemetry, Token-Conservation Analytics & Operational Reporting — vNext.10.2 Phase 9
+
+- Added a versioned Job/Mission execution telemetry report derived from
+  durable WorkUnit, routing, attempt, research, cooldown, verification,
+  closure, and autonomy facts. Reports keep unknown usage as `null`, expose
+  coverage, deduplicate replayed attempt IDs, redact diagnostics, and remain
+  observational rather than granting authority or completion.
+- Added formal Secondary success/repair/fallback funnels,
+  `StrongBuilderAvoidanceRatio` with Strong evaluator separation, Strong
+  direct/fallback attribution, research avoidance/reuse and lifecycle
+  breakdowns, cooldown productivity/avoidable-idle diagnostics, restart and
+  candidate reuse accounting, completed-work redo detection, and a compact
+  Phase 10-compatible qualification summary.
+- Added `specbridge report job <jobId>` with JSON and verbose formats plus the
+  read-only MCP `job_report` tool. CLI reports persist under
+  `.specbridge/reports/` by default; MCP derives the same contract without
+  writing runtime state.
+- Added deterministic mixed-compute, Strong-only replay, research, cooldown,
+  restart/deduplication, token coverage, intervention, reliability, old-Job,
+  corruption, CLI, and MCP qualification coverage, and regenerated the public
+  contracts, tool reference, and Claude/Codex plugin bundles.
+
 ### Subscription Cooldown Continuation & Compute Survival — vNext.10.2 Phase 8
 
 - Added durable, resource-scoped Strong subscription cooldown state at the

diff --git a/contracts/cli-commands.json b/contracts/cli-commands.json
@@ -761,6 +761,21 @@
           }
         }
       },
+      "report": {
+        "options": [
+          "--help"
+        ],
+        "subcommands": {
+          "job": {
+            "options": [
+              "--help",
+              "--json",
+              "--no-persist",
+              "--verbose"
+            ]
+          }
+        }
+      },
       "research": {
         "options": [
           "--help"

diff --git a/contracts/mcp-contract.json b/contracts/mcp-contract.json
@@ -27,6 +27,7 @@
     "job_cancel",
     "job_list",
     "job_read",
+    "job_report",
     "mission_answer",
     "mission_assess",
     "mission_begin",

diff --git a/docs/README.md b/docs/README.md
@@ -129,6 +129,9 @@ the intent in the evening, run one command, read the report in the morning.
   self-repair that cannot weaken its own constraints.
 - [Autonomy telemetry](autonomy/telemetry.md) — the honest report, where an
   unknown is `null` and never zero.
+- [Execution telemetry and token conservation](autonomy/execution-telemetry.md)
+  — versioned Phase 9 Job/WorkUnit routing, research, cooldown, reliability,
+  verification, and closure analytics without new authority.
 - [Zero-touch certification](autonomy/certification.md) — the sixteen-fault
   matrix, and the three defects it found.
 - [Operator setup](autonomy/operator-setup.md) — the evening checklist.
@@ -243,7 +246,7 @@ the intent in the evening, run one command, read the report in the morning.
 ## MCP & frontend plugins
 
 - [MCP server](mcp-server.md) — the local stdio server.
-- [MCP tool reference](mcp/tool-reference.md) — all 77 tools (generated).
+- [MCP tool reference](mcp/tool-reference.md) — all 78 tools (generated).
 - [MCP tools](mcp-tools.md) · [resources](mcp-resources.md) ·
   [prompts](mcp-prompts.md) · [CLI/MCP parity](cli-mcp-parity.md).
 - [Claude Code integration](claude-code-integration.md) — both directions.

diff --git a/docs/autonomy/execution-telemetry.md b/docs/autonomy/execution-telemetry.md
@@ -0,0 +1,93 @@
+# Execution telemetry and token-conservation reports
+
+Phase 9 turns durable Job, WorkUnit, routing, research, cooldown,
+verification, closure, and autonomy records into one versioned operational
+report. The report is observational: it cannot approve work, change routing,
+grant authority, or make a Job complete.
+
+## Read a report
+
+```bash
+specbridge report job <jobId>
+specbridge report job <jobId> --verbose
+specbridge report job <jobId> --json
+```
+
+The human view is concise by default. `--verbose` adds bounded WorkUnit
+accounting and the cooldown timeline. The JSON view carries
+`schemaVersion: "1.0.0"` and, unless `--no-persist` is used, is saved at
+`.specbridge/reports/job-<jobId>-telemetry.json`.
+
+MCP clients can call the read-only `job_report` tool. It derives the same
+report without persisting a file.
+
+## Formal metric definitions
+
+A ratio is represented as `{ numerator, denominator, value }`. `value` is
+`null` when the denominator is zero or the underlying durable facts do not
+exist. Unknown measurements are never reported as zero.
+
+| Metric | Numerator | Denominator | Exclusions and unknown semantics |
+| --- | --- | --- | --- |
+| `StrongBuilderAvoidanceRatio` | Eligible, completed implementation WorkUnits with no `STRONG` or `STRONG_FALLBACK` builder attempt | Eligible, completed implementation WorkUnits | Excludes `STRONG_REQUIRED`, cancelled, authority-pending, research-pending, context-pending, and incomplete work. A Strong evaluator is not a Strong Builder. `null` when no eligible implementation WorkUnit completed. |
+| `SecondaryInitialSuccessRate` | WorkUnits whose initial `SECONDARY` attempt passed without repair | WorkUnits with an initial `SECONDARY` attempt | `null` when Secondary was never initially attempted. Replayed attempt IDs are counted once. |
+| `SecondaryRepairRecoveryRate` | WorkUnits whose `SECONDARY_REPAIR` path passed | WorkUnits that received a `SECONDARY_REPAIR` attempt | `null` when repair was never attempted. |
+| `SecondaryToStrongFallbackRate` | Secondary-attempted WorkUnits that later used `STRONG_FALLBACK` | WorkUnits with any Secondary implementation attempt | `null` when Secondary was never attempted. |
+| `ResearchAvoidanceRatio` | Research-gate considerations resolved without a new QUICK or DEEP provider call | Research-gate considerations | `null` where gate aggregates are unavailable. The report labels existing aggregate gate telemetry as `WORKSPACE`; it does not pretend it is Job-scoped. Job-scoped provider records remain Job-scoped. |
+| `ResearchReuseRate` | Reused prior research reports | Reused reports plus new provider calls | `null` when neither reuse nor a new call occurred. |
+| `UsefulWorkDuringSubscriptionCooldown` | Distinct WorkUnits durably recorded as completed during a Strong subscription cooldown | Not a ratio | Includes productive Secondary and research progress while Strong-required work waits. Duplicate WorkUnit IDs within the same objective are counted once. |
+| `ZeroTouchAfterSeal` | `true` only when `humanInterventionsAfterSeal` is zero | Not a ratio | Correct authority escalation is reported separately. A missing seal boundary is diagnosed by the underlying autonomy telemetry; it is never inferred from a successful outcome. |
+| `CompletedWorkRedoCount` | Implementation attempts that began after the same WorkUnit had already been durably integrated | Not a ratio | Resource waits and replayed copies of the same durable attempt ID are excluded. Candidate reuse and restart recovery are reported separately. |
+
+## Work accounting
+
+Every current WorkUnit belongs to exactly one reporting category:
+`completed`, `failed`, `cancelled`, `waiting`, `not-ready`,
+`human-authority-pending`, `research-pending`, or `context-pending`. This is
+reporting classification only and does not alter the WorkUnit state machine.
+
+The same accounting is rolled up per Objective and for the whole Job;
+`missionId` links the Job report to its sealed Mission when that durable
+binding exists. Secondary eligibility reports both the eligible total and an
+explicit ineligible total, with Strong-required, research, authority,
+context, and not-ready reasons kept separate.
+
+Implementation builder calls and evaluator calls are separate populations.
+`STRONG` and `STRONG_FALLBACK` count as Strong Builder calls;
+non-local `EVALUATOR` ledger entries count only as Strong evaluator calls.
+
+## Token and research coverage
+
+Provider-reported input and output tokens are summed only when present.
+`knownTokens` is the sum of observed components; `completeTokens` is `null`
+unless every attempt reported both components. Coverage states how many
+attempts reported any, input, and output usage. Research cost and duration
+follow the same rule: absence is `null`, not free or instantaneous.
+
+Research reports expose `JOB`, `WORKSPACE`, or `NONE` scope. Lifecycle
+breakdowns cover conversation, spec draft, intake decision, and runtime
+investigation. Workspace-scoped legacy aggregates are explicitly labeled so
+cross-Job evidence cannot be mistaken for a Job-only total.
+
+## Provenance, safety, and compatibility
+
+The report records Job and objective graph revisions, event and execution
+ledger watermarks, seal references, the current non-secret strategy summary,
+and whether the current runtime policy differs from the sealed binding.
+Diagnostic text is bounded and redacted. Prompts, transcripts, raw provider
+payloads, environment secrets, and credentials are never report fields.
+
+Older Jobs remain reportable. Missing Phase 7/8, research, seal, or closure
+records produce empty or `null` sections plus diagnostics rather than
+invented facts. Inputs and output arrays are bounded, and duplicate durable
+attempt IDs are deterministically deduplicated.
+
+## Comparing routing strategies
+
+Efficiency claims require comparable verification and closure outcomes.
+The JSON report can carry an explicitly identified observed or qualification
+baseline; it reports whether outcomes match before presenting Strong Builder
+call reduction. `correctnessEqual` remains false when trusted verification or
+closure evidence is unavailable, even when both missing values match. This is
+evidence for qualification and operations, not a learned-routing or
+production-readiness claim.
diff --git a/docs/autonomy/telemetry.md b/docs/autonomy/telemetry.md
@@ -89,6 +89,11 @@ first process restart — and process restarts are the normal case here.
 specbridge autonomy report <jobId>
 ```
 
+For the Phase 9 Job/WorkUnit routing, research, cooldown, token-conservation,
+verification, and closure report, see
+[Execution telemetry and token conservation](execution-telemetry.md) and run
+`specbridge report job <jobId>`.
+
 ```
 Autonomy
   ✓ humanInterventionsAfterSeal: 0

diff --git a/docs/mcp/tool-reference.md b/docs/mcp/tool-reference.md
@@ -7,7 +7,7 @@ Generated from the authoritative registries of the `specbridge` MCP server
 (version 1.1.0). Tool names, resource URI templates, and prompt
 names are stable contracts — see docs/stability/public-contracts.md.
 
-## Tools (77)
+## Tools (78)
 
 | Tool | Access | Summary |
 | --- | --- | --- |
@@ -22,6 +22,7 @@ names are stable contracts — see docs/stability/public-contracts.md.
 | `job_cancel` | write | Cancel a job (final, idempotent, evidence preserved) |
 | `job_list` | read-only | List long-running orchestration jobs |
 | `job_read` | read-only | One job in depth: graph, attempts, questions, checkpoint |
+| `job_report` | read-only | Durable Job operational and token-conservation report |
 | `mission_answer` | write | Record the user’s answer to one discovery question |
 | `mission_assess` | write | Record a governed structured discovery assessment |
 | `mission_begin` | write | Begin Mission Discovery from a product direction |

diff --git a/integrations/claude-code-plugin/specbridge/dist/checksums.json b/integrations/claude-code-plugin/specbridge/dist/checksums.json
@@ -7,12 +7,12 @@
       "bytes": 155994
     },
     "cli.cjs": {
-      "sha256": "9b6d58aef80578e88c81a1c48606cb13ac0d02d41d9acccc7e3cc67348c1c698",
-      "bytes": 5739368
+      "sha256": "91d51d28006410e893ba0f19b1989e0e170a4c2820bf3fd2770a84686777fe2c",
+      "bytes": 5814917
     },
     "mcp-server.cjs": {
-      "sha256": "f096d9dff325c106c15334e6c67baef15771789e5aa2ddd06858894987ae9b4c",
-      "bytes": 3669226
+      "sha256": "0cf0895d3af59c3de92fbf899965f2c093145e1f5aa3bf23846ec59e9a5568d6",
+      "bytes": 3758229
     }
   }
 }