Add cross-platform WGO support and reproducible audit cost controls

.claude-plugin/plugin.json

@@ -0,0 +1,22 @@
+{
+  "$schema": "https://json.schemastore.org/claude-code-plugin.json",
+  "name": "wgo",
+  "displayName": "Whats.Going.On.",
+  "version": "0.2.0+claude.20260804025205",
+  "description": "Evidence-led startup and SMB audits with decision-grade reports, detailed control artifacts, and optional draft operator aids.",
+  "author": {
+    "name": "Patrick Drolet",
+    "email": "patrick@drolet.name",
+    "url": "https://github.com/Patrick-Loves-Espresso"
+  },
+  "homepage": "https://wgo-audit.com",
+  "repository": "https://github.com/wgo-audit/code",
+  "license": "MIT",
+  "keywords": [
+    "audit",
+    "startup",
+    "smb",
+    "due-diligence",
+    "repo-analysis"
+  ]
+}

.codex-plugin/plugin.json

@@ -5,10 +5,10 @@
   "author": {
     "name": "Patrick Drolet",
     "email": "patrick@drolet.name",
-    "url": "https://github.com/PatD42"
+    "url": "https://github.com/Patrick-Loves-Espresso"
   },
-  "homepage": "https://github.com/PatD42/whats-going-on",
-  "repository": "https://github.com/PatD42/whats-going-on",
+  "homepage": "https://wgo-audit.com",
+  "repository": "https://github.com/wgo-audit/code",
   "license": "MIT",
   "keywords": [
     "audit",
@@ -28,7 +28,7 @@
       "Interactive",
       "Write"
     ],
-    "websiteURL": "https://github.com/PatD42/whats-going-on",
+    "websiteURL": "https://wgo-audit.com",
     "defaultPrompt": [
       "Run wgo:onboard.",
       "Run wgo:audit product-value.",

README.md

@@ -8,7 +8,7 @@ When key developers leave, a vendor handoff goes sideways, costs stop making sen
 
 Whats.Going.On. (WGO) is an open-source, prompt-first audit plugin for startups and SMBs. It guides an auditor from a plain-language concern to evidence-backed findings, owned decisions, and reports written for the people who must act on them.
 
-**WGO runs locally with Codex or Claude. There is no hosted service, database, dashboard, or proprietary scoring model.**
+**WGO runs locally with Codex, Claude, or OpenCode. There is no hosted service, database, dashboard, or proprietary scoring model.**
 
 > WGO is not another static analyzer. It is a semi-structured audit lead that helps you ask the right questions, inspect the right evidence, preserve uncertainty, and explain the result clearly.
 
@@ -43,29 +43,44 @@ Clone WGO, then install it into the project you want to audit.
 macOS or Linux:
 
 ```bash
-git clone https://github.com/PatD42/whats-going-on.git
+git clone https://github.com/wgo-audit/code.git whats-going-on
 cd whats-going-on
 ./install.sh /path/to/project
 ```
 
 Windows:
 
 ```bat
-git clone https://github.com/PatD42/whats-going-on.git
+git clone https://github.com/wgo-audit/code.git whats-going-on
 cd whats-going-on
 install.bat C:\path\to\project
 ```
 
-Open the target project in Codex or Claude and start onboarding:
+Open the target project in Codex, Claude, or OpenCode and start onboarding:
 
 ```text
 # Codex
 wgo:onboard
 
 # Claude
-/wgo_onboard
+/wgo:onboard
+
+# OpenCode
+/wgo-onboard
 ```
 
+The installer creates a native Codex plugin at `plugins/wgo` and a native
+Claude plugin at `.claude/skills/wgo-claude`. It renders the same canonical
+skill for OpenCode at `.opencode/skills/wgo` and adds thin command wrappers at
+`.opencode/commands/wgo-*.md`. All three providers are rendered from the same
+canonical skill and command files; provider-specific frontmatter is filtered
+from installed copies, so the workflow cannot drift between maintained
+implementations.
+On first use, accept Claude's workspace-trust prompt; if Claude was already
+open during installation, run `/reload-plugins` or reopen the project.
+Restart OpenCode after installation. Until OpenCode indexes newly installed
+skill files, its command wrappers load `.opencode/skills/wgo/SKILL.md` directly.
+
 For a first audit, WGO confirms the company, product, business concern,
 additional evidence, and reviewers. It creates
 `_whats-going-on-YYYYMMDD/`, dated when onboarding starts, only after approval.
@@ -113,7 +128,7 @@ flowchart TB
 1. **Onboard:** establish the product, mandate, evidence boundary, success criteria, and selected reviewers. WGO also discovers project-local reviewer extensions, but never selects one without approval.
 2. **Audit:** inspect one reviewer or run all selected reviewers in coordinator-defined dependency waves; register evidence, ask material questions, and update shared controls.
 3. **Check status:** see completed work, open proof, access blockers, risks, decisions, and remaining reviewers.
-4. **Summarize:** reconcile the whole audit and create decision-grade reports for each audience.
+4. **Summarize:** reconcile the whole audit, create decision-grade reports for each audience, and produce a reconciled API-equivalent audit cost estimate from Codex, Claude, or OpenCode's own session records.
 5. **Operationalize when justified:** after a completed audit, explicitly request draft operator aids. WGO will not execute them or change a system.
 
 Run a named reviewer when you want focused control. With no reviewer argument,
@@ -122,13 +137,14 @@ that do not depend on each other.
 
 ## Commands
 
-| Purpose | Codex | Claude |
-|---|---|---|
-| Start, improve, or compare an audit | `wgo:onboard` with `[compare\|blind-compare] [YYYYMMDD]` | `/wgo_onboard` with `[compare\|blind-compare] [YYYYMMDD]` |
-| Run one reviewer or all selected reviewers | `wgo:audit [reviewer-id\|all]` | `/wgo_audit [reviewer-id\|all]` |
-| Show truthful progress and blockers | `wgo:status` | `/wgo_status` |
-| Reconcile findings and generate reports | `wgo:summarize` | `/wgo_summarize` |
-| Draft operator aids after approved synthesis | `wgo:operationalize` | `/wgo_operationalize` |
+| Purpose | Codex | Claude | OpenCode |
+|---|---|---|---|
+| Start, improve, or compare an audit | `wgo:onboard` with `[compare\|blind-compare] [YYYYMMDD]` | `/wgo:onboard` with `[compare\|blind-compare] [YYYYMMDD]` | `/wgo-onboard` with `[compare\|blind-compare] [YYYYMMDD]` |
+| Run one reviewer or all selected reviewers | `wgo:audit [reviewer-id\|all]` | `/wgo:audit [reviewer-id\|all]` | `/wgo-audit [reviewer-id\|all]` |
+| Show truthful progress and blockers | `wgo:status` | `/wgo:status` | `/wgo-status` |
+| Reconcile findings and generate reports | `wgo:summarize` | `/wgo:summarize` | `/wgo-summarize` |
+| Produce a reconciled API-equivalent audit cost estimate | `wgo:cost` | `/wgo:cost` | `/wgo-cost` |
+| Draft operator aids after approved synthesis | `wgo:operationalize` | `/wgo:operationalize` | `/wgo-operationalize` |
 
 The audit always concerns the full project in the current folder, including all
 repository subfolders.
@@ -198,6 +214,19 @@ An audit should not silently turn into operational work. `wgo:operationalize` is
 2. The auditor explicitly requests operationalization.
 3. Creating local audit artifacts is permitted.
 
+When operationalization completes, WGO refreshes `controls/cost-estimate.md`
+through the operator-aid phase. It preserves the earlier audit-only cost
+manifest and excludes both cost-calculation passes from the estimate.
+
+Cost evidence is provider-specific rather than translated into a synthetic
+common log format. Codex uses rollout JSONL request usage, Claude uses its root and
+subagent JSONL transcripts with Anthropic's dated API rate card, and OpenCode
+uses validated session exports with their recorded per-message provider cost.
+Every platform freezes the included session graph before two independent
+calculations. Missing provenance, malformed exports, unknown rates, and
+unexplained zero-cost OpenCode models produce `Unreconciled` or `unpriced`
+results instead of a guessed `$0.00`.
+
 It produces a focused four-part operating packet. These are complementary
 operating perspectives, not variants of one generic runbook:
 
@@ -240,17 +269,18 @@ expansion. It does not quietly treat the referenced material as reviewed.
 ## Improve Or Compare An Audit
 
 WGO audit state is local, readable Markdown rather than model-specific memory.
-An auditor can begin an audit with Codex and later improve the same dated audit
-root with Claude, or the reverse. Plain `wgo:onboard` reopens the newest root
-read-write, displays its configuration, and asks whether anything needs
-updating.
-
-`wgo:onboard compare [YYYYMMDD]` creates today's root and performs a targeted,
-read-only comparison against the named completed baseline or the latest
-completed audit. `wgo:onboard blind-compare [YYYYMMDD]` performs a full audit
-without exposing baseline findings, then compares the completed audits. Both
-comparison modes report reviewer-version differences and, when versions differ,
-require the auditor to accept the installed versions before proceeding.
+An auditor can begin an audit with Codex, Claude, or OpenCode and later improve
+the same dated audit root with either of the other providers. The provider's
+plain onboarding command reopens the newest root read-write, displays its
+configuration, and asks whether anything needs updating.
+
+Adding `compare [YYYYMMDD]` to the provider's onboarding command creates today's
+root and performs a targeted, read-only comparison against the named completed
+baseline or the latest completed audit. Adding `blind-compare [YYYYMMDD]`
+performs a full audit without exposing baseline findings, then compares the
+completed audits. Both comparison modes report reviewer-version differences
+and, when versions differ, require the auditor to accept the installed versions
+before proceeding.
 
 The result is cumulative, but not blindly append-only:
 

commands/audit.md

@@ -3,15 +3,18 @@ name: audit
 description: Run one selected Whats.Going.On. reviewer, or all selected reviewers in dependency waves.
 args: "[reviewer-id|all]"
 skills: wgo
+argument-hint: "[reviewer-id|all]"
+disable-model-invocation: true
 ---
 
 # /audit
 
-Codex users may invoke this as `wgo:audit <reviewer-id>`, `wgo:audit all`, or
-`wgo:audit`.
+Invoke this as `wgo:audit <reviewer-id>`/`wgo:audit all` in Codex,
+`/wgo:audit <reviewer-id>`/`/wgo:audit all` in Claude, or
+`/wgo-audit <reviewer-id>`/`/wgo-audit all` in OpenCode.
 
-Resolve the newest dated audit root with
-`skills/wgo/references/common/audit-root.md`. Read and write only that root.
+Load and use the WGO skill. Resolve the newest dated audit root with its
+`references/common/audit-root.md`. Read and write only that root.
 
 ## One Reviewer
 
@@ -45,7 +48,8 @@ then:
    insights when they meet the shared standard, and compact handoff; then update
    the checklist and material shared open items.
 
-If this completes the last selected reviewer, automatically run `wgo:summarize`.
+If this completes the last selected reviewer, pass the delegated-task lifecycle
+gate below, then automatically run `wgo:summarize`.
 
 ## All Selected Reviewers
 
@@ -83,8 +87,9 @@ checklist. At the end of the wave, surface a qualifying auditor question before
 starting dependent work: ask one exact question, wait for its answer, record it
 in the brief and its existing `decision-needed` item, then ask another only if
 it remains material. Do not let `wgo:audit all` defer such a question merely to
-keep the run uninterrupted. Then validate the audit root and start the next
-wave. Once all selected reviewers are complete, automatically run
+keep the run uninterrupted. Then pass the delegated-task lifecycle gate,
+validate the audit root, and start the next wave. Once all selected reviewers
+are complete, pass the gate again and automatically run
 `wgo:summarize`. Under `complete-missing`, if no reviewer work was missing and
 the existing synthesis already covers the unchanged selected reviewer set, do
 not rewrite it; report that nothing was missing. Under `rerun-all`, always rerun
@@ -96,5 +101,18 @@ to reassess them. They do not search for unrelated new findings. For
 `blind-compare`, do not give reviewers any baseline content or prior
 identifiers.
 
+## Delegated-Task Lifecycle Gate
+
+Before ending a reviewer, advancing a dependency wave, or starting synthesis,
+recursively inventory every WGO task spawned during the current audit from the
+recorded collaboration and task-lifecycle provenance. Require exactly one
+terminal outcome (`completed`, `failed`, `cancelled`, or `interrupted`) for
+every recorded start. Do not infer closure from a report, checklist state,
+message, or worker disappearance. Wait for a running task; if a start remains
+open, has multiple terminal outcomes, or cannot be correlated unambiguously,
+stop the transition and identify the exact task/session. A failed, cancelled,
+or interrupted task is lifecycle-closed but still follows the existing
+dependency and audit-limitation rules.
+
 Do not mark work complete merely because a report exists. A bounded,
 completed-with-open-verification conclusion is valid when its limits are clear.

commands/cost.md

@@ -0,0 +1,36 @@
+---
+name: cost
+description: Produce a reconciled API-equivalent cost estimate for a completed Whats.Going.On. audit.
+skills: wgo
+disable-model-invocation: true
+---
+
+# /cost
+
+Invoke this as `wgo:cost` in Codex, `/wgo:cost` in Claude, or `/wgo-cost` in
+OpenCode.
+
+Load and use the WGO skill. Resolve the newest dated audit root with its
+`references/common/audit-root.md`, then read the audit platform recorded in
+`audit-brief.md` and run exactly one matching workflow:
+
+- Codex: `references/common/cost-estimation.md`;
+- Claude: `references/common/cost-estimation-claude.md`; or
+- OpenCode: `references/common/cost-estimation-opencode.md`.
+
+Do not inspect a provider's session store or apply its usage schema to an audit
+recorded on another platform. If the platform is missing, ambiguous, or
+unsupported, write an `Unreconciled` cost control naming that limitation rather
+than guessing.
+
+This is a read-only analysis of accessible provider-native session records that
+writes `<audit-root>/controls/cost-estimate.md` and its frozen calculation
+evidence.
+Do not install a package, invoke `ccusage`, add a helper program, or use an
+OS-specific binary.
+The active-platform coordinator discovers the audit descendants from recorded
+collaboration and task-lifecycle provenance, freezes the manifest before any
+calculation, and runs exactly two independent calculations over that unchanged
+manifest using the worker configuration in the selected provider workflow.
+Never identify an included session solely by date, CWD, folder, or model; those
+may corroborate recorded provenance only.

commands/onboard.md

@@ -3,12 +3,15 @@ name: onboard
 description: Design or resume a Whats.Going.On. audit for the current project folder.
 args: "[compare|blind-compare] [YYYYMMDD]"
 skills: wgo
+argument-hint: "[compare|blind-compare] [YYYYMMDD]"
+disable-model-invocation: true
 ---
 
 # /onboard
 
-Resolve audit roots with `skills/wgo/references/common/audit-root.md`, then run
-`skills/wgo/references/common/onboarding.md`.
+Load and use the WGO skill. Resolve audit roots with its
+`references/common/audit-root.md`, then run its
+`references/common/onboarding.md`.
 
 The audit covers the full current project folder at detailed depth. Quietly
 perform read-only discovery and confirm company/product context. When a prior

commands/operationalize.md

@@ -2,16 +2,19 @@
 name: operationalize
 description: Turn an approved synthesis into explicitly untested, source-linked operator aids without executing procedures or changing systems.
 skills: wgo
+disable-model-invocation: true
 ---
 
 # /operationalize
 
-Codex users may invoke this as `wgo:operationalize`.
+Invoke this as `wgo:operationalize` in Codex, `/wgo:operationalize` in Claude,
+or `/wgo-operationalize` in OpenCode.
 
-Resolve the newest dated audit root with
-`skills/wgo/references/common/audit-root.md`.
+Load and use the WGO skill. Resolve the newest dated audit root with its
+`references/common/audit-root.md`.
 
-Run the complete workflow in `skills/wgo/references/common/operationalization.md`.
+Run the complete workflow in its
+`references/common/operationalization.md`.
 
 Hard rules:
 
@@ -25,3 +28,9 @@ Hard rules:
 - Create an operating-control strategy only when the completed controls need an explicit stabilization and authority model; it is a derived decision aid, not a second backlog or a runbook.
 - Create the four-part transition packet at `operator-aids/`: `replacement-maintainer`, `recovery`, `observability`, and `iam-and-credential-control`. Keep them separate, cross-linked, and evidence-bounded.
 - Do not create a fixed-size 72-hour mission board. Generate the execution-priority taxonomy instead.
+- After the packet is complete, emit the operationalization-complete cutoff and
+  rerun the Codex, Claude, or OpenCode cost workflow selected from the platform
+  recorded in `audit-brief.md`. Update
+  `controls/cost-estimate.md` through operationalization without modifying the
+  frozen audit-only cost evidence or charging either cost-calculation phase to
+  the audit.

commands/status.md

@@ -2,14 +2,16 @@
 name: status
 description: Show truthful Whats.Going.On. audit progress, blockers, and material open items.
 skills: wgo
+disable-model-invocation: true
 ---
 
 # /status
 
-Codex users may invoke this as `wgo:status`.
+Invoke this as `wgo:status` in Codex, `/wgo:status` in Claude, or
+`/wgo-status` in OpenCode.
 
-Resolve the newest dated audit root with
-`skills/wgo/references/common/audit-root.md`. Read, without modifying:
+Load and use the WGO skill. Resolve the newest dated audit root with its
+`references/common/audit-root.md`. Read, without modifying:
 
 - `<audit-root>/audit-checklist.md`
 - `<audit-root>/controls/open-items.md`

commands/summarize.md

@@ -2,16 +2,18 @@
 name: summarize
 description: Reconcile completed Whats.Going.On. reviewers and create decision-grade audience reports.
 skills: wgo
+disable-model-invocation: true
 ---
 
 # /summarize
 
-Codex users may invoke this as `wgo:summarize`.
+Invoke this as `wgo:summarize` in Codex, `/wgo:summarize` in Claude, or
+`/wgo-summarize` in OpenCode.
 
-Resolve the newest dated audit root with
-`skills/wgo/references/common/audit-root.md`.
+Load and use the WGO skill. Resolve the newest dated audit root with its
+`references/common/audit-root.md`.
 
-Run `skills/wgo/references/common/synthesis.md`. Read completed
+Run its `references/common/synthesis.md`. Read completed
 reviewer reports, their handoffs, and linked evidence/artifacts directly.
 Reconcile before drafting and preserve source-access limits, uncertainty, and
 evidence cutoff.
@@ -21,5 +23,13 @@ and `technical-lead-notes.md` at the audit root. Include an owner-assigned
 30–90 day section only where evidence supports it. Do not create a separate
 reconciliation command, decision queue, or action backlog.
 
+After synthesis validation succeeds, run the complete `wgo:cost` phase before
+asking about operationalization. It must discover and freeze the audit-session
+manifest itself, select the provider workflow from `audit-brief.md`, use its two
+independent verification passes over that manifest, and write
+`controls/cost-estimate.md`. Link that control from `index.md` and the three
+audience reports. A cost result may be `Unreconciled`; do not suppress it or
+replace it with a precise total.
+
 After the summary is complete, ask exactly: `Should I proceed with
 wgo:operationalize?` Do not draft operator aids until the auditor answers yes.