@webpresso/opencode-plugin 3.1.20 → 3.1.22

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@webpresso/opencode-plugin",
-  "version": "3.1.20",
+  "version": "3.1.22",
   "private": false,
   "description": "OpenCode plugin adapter package for Webpresso agent-kit.",
   "homepage": "https://github.com/webpresso/agent-kit#readme",

package/plugin-skill-ownership.json

@@ -2,7 +2,7 @@
   "schemaVersion": 1,
   "host": "opencode",
   "packageName": "@webpresso/opencode-plugin",
-  "packageVersion": "3.1.20",
+  "packageVersion": "3.1.22",
   "runtimeDirs": [],
   "skills": {
     "ai-deslop": {
@@ -21,10 +21,10 @@
       "digest": "sha256:21fd24862e7f7c8a1feadea6bc376492b136b99fc381cf1b8927c9db80615431"
     },
     "claude": {
-      "digest": "sha256:dec5cc89c31485bab6defea919c22e9a9c50c530d8fc4ef4f5be5080b5d91b75"
+      "digest": "sha256:bca8cc1311bb5a21aa7367ba5c71c3ef8269f29fcb82ec70df24e4c3f7ea50ac"
     },
     "codex": {
-      "digest": "sha256:98187c2845b5cd88ba9eb9f682d16fec648932a084b3d1cebd066945b1fb6fdf"
+      "digest": "sha256:acc5817fa9186891cdf9a17a5217fc89a98d7ac04620c363a715b312158bbcad"
     },
     "deep-interview": {
       "digest": "sha256:e80d7e12e486397103108bb473e98e8e8656519770fa20701f8b3e9e66991939"
@@ -41,8 +41,11 @@
     "fix": {
       "digest": "sha256:647aa94890e06b38b08e6cc12cd219343e963b4d3b469c349610b3a33f559dd5"
     },
+    "grok": {
+      "digest": "sha256:663b47c6abac5be939480e28a32029156f2b1ffa75bfe42c6b1edbbb52573a55"
+    },
     "hooks-doctor": {
-      "digest": "sha256:3bfb0af64ff6796c56d51e4cd2bede7fb24c869f89b292546ea13ad2f5d3bb73"
+      "digest": "sha256:d3e000980b9de7834233b98321c78000a37d38e303321186a0fc3268f6f51e69"
     },
     "investigate": {
       "digest": "sha256:2051bccb2daf97ee91a8c15c7ef5bc7cc705d49beb4cbc7c8f6c63f4acda88a2"
@@ -51,7 +54,7 @@
       "digest": "sha256:a80fbacc765e8b886437b9fdf7dd71e1ceb1b98b939078ec90fc88d06a8b4c69"
     },
     "opencode-go": {
-      "digest": "sha256:8bbfad8273c526c563ceef3c8b98efbc60c3b24dc6b0a2a25d8efd9e83788860"
+      "digest": "sha256:b36683da200b59fc5a8cceb8da34b7b29613680c4e52c4f4aaae03ff3acdd269"
     },
     "plan-ceo-review": {
       "digest": "sha256:4d426158518dd71f8998770e8d4037bcbfa3c65d7265a93a4a78a9b015ba881e"
@@ -87,7 +90,7 @@
       "digest": "sha256:dc6d3b2bd48a494d7a5cd30ad9c6c9999989b9920ca06ccd1f86a7cfaff60d54"
     },
     "verify": {
-      "digest": "sha256:2964a54ef0f9db389d3ef5fa779fa2f095c9f8df47f19f3e4947d2403cdbf156"
+      "digest": "sha256:9d9faa184b310d46d5a266ebe46bbf0819df526035dc10a6c13475097e3a284b"
     }
   }
 }

package/skills/claude/SKILL.md

@@ -57,7 +57,13 @@ Claude's managed native subagent lifecycle does not provide a reliable parent-li
 
 Use single-file / single-question first for any non-trivial diff. Do not send a whole PR unless it already fits within the bounded payload below.
 
-**Model policy:** default to the moving `opus` alias via `CLAUDE_REVIEW_MODEL=${CLAUDE_REVIEW_MODEL:-opus}`. A caller may explicitly override `CLAUDE_REVIEW_MODEL` (for example to `fable` for a lighter advisory pass); honor the requested model. However, any review used as blueprint **promotion or completion** approval evidence (draft→planned, or planned/in-progress→completed) MUST run with Claude Opus and record `"model": "opus"`.
+**Model policy:** leave `CLAUDE_REVIEW_MODEL` unset unless the user requested a
+specific Claude model (for example `fable` for a lighter advisory pass). When it
+is unset, `wp review run` lets the Claude CLI choose its own default by omitting
+`--model`. The blueprint promotion/completion approval gate
+(`catalog/agent/rules/pre-implementation.md`) validates reviewer identity
+(`claude`/`codex`/`grok`/an OpenCode-Go model), not the reviewed model string, so do not
+claim the gate rejects a different Claude model.
 
 #### Bounded prompt payload
 
@@ -94,8 +100,17 @@ TARGET_FILE=${TARGET_FILE:?set TARGET_FILE to one changed file}
 
 Run both the availability probe and the full review through the typed `wp review run` owner. Skills must not embed subprocess supervision. The runtime consumes Claude's streaming JSON events, advances its idle clock only on monotonic semantic progress, and has no total wall-clock cutoff.
 
+The artifact root contains private runtime diagnostics, not a recorded
+transcript or committable approval evidence. Direct `wp review run` output is
+advisory. Only `wp review gate` may publish a bounded, target-bound artifact and
+ledger event for blueprint promotion or completion.
+
 ```bash
-CLAUDE_REVIEW_MODEL=${CLAUDE_REVIEW_MODEL:-opus}
+CLAUDE_REVIEW_MODEL=${CLAUDE_REVIEW_MODEL:-}
+CLAUDE_MODEL_ARGS=()
+if [ -n "$CLAUDE_REVIEW_MODEL" ]; then
+  CLAUDE_MODEL_ARGS=(--model "$CLAUDE_REVIEW_MODEL")
+fi
 CLAUDE_REVIEW_EFFORT=${CLAUDE_REVIEW_EFFORT:-medium}
 CLAUDE_REVIEW_PROBE_IDLE_SECONDS=${CLAUDE_REVIEW_PROBE_IDLE_SECONDS:-180}
 CLAUDE_REVIEW_IDLE_SECONDS=${CLAUDE_REVIEW_IDLE_SECONDS:-180}
@@ -108,7 +123,7 @@ CLAUDE_REVIEW_CODE=0
 if wp review run \
   --provider claude \
   --prompt-file "$PROBE_FILE" \
-  --model "$CLAUDE_REVIEW_MODEL" \
+  "${CLAUDE_MODEL_ARGS[@]}" \
   --effort "$CLAUDE_REVIEW_EFFORT" \
   --stage probe \
   --expected-marker CLAUDE_OK \
@@ -117,7 +132,7 @@ if wp review run \
   wp review run \
     --provider claude \
     --prompt-file "$PROMPT_FILE" \
-    --model "$CLAUDE_REVIEW_MODEL" \
+    "${CLAUDE_MODEL_ARGS[@]}" \
     --effort "$CLAUDE_REVIEW_EFFORT" \
     --stage review \
     --artifact-root "$CLAUDE_REVIEW_ARTIFACT_ROOT" \
@@ -141,7 +156,7 @@ if [ -n "$DASH_HELPER_RUN_ID" ]; then
 fi
 ```
 
-Treat `true-idle`, `protocol-unsupported`, provider failure, abort, spawn failure, or artifact failure as unavailable review evidence. Never replace them with a static timeout, arbitrary byte-growth heartbeat, or buffered-output fallback.
+Treat `true-idle`, `protocol-unsupported`, provider failure, abort, spawn failure, or artifact failure as an unavailable advisory result. Never replace them with a static timeout, arbitrary byte-growth heartbeat, or buffered-output fallback.
 
 #### Retry policy
 

package/skills/codex/SKILL.md

@@ -53,20 +53,26 @@ Codex native subagent events are not currently parent-linked in the managed life
 ## Options
 
 Every mode runs through the `wp review run` runtime and honors these environment
-overrides (defaults shown). Set them inline, e.g.
+overrides. Set them inline, e.g.
 `CODEX_REVIEW_MODEL=gpt-5.6-sol CODEX_REVIEW_EFFORT=high /codex`.
 
 - `CODEX_REVIEW_MODEL` — the Codex model. Defaults to the `model = "..."` in the
-  active Codex config (`${CODEX_HOME:-~/.codex}/config.toml`), falling back to
-  `gpt-5.4`. It MUST be a model the current login is provisioned for: ChatGPT-account
+  active Codex config (`${CODEX_HOME:-~/.codex}/config.toml`); if neither the
+  environment nor config names a model, `wp review run` lets the Codex CLI choose
+  its own default. It MUST be a model the current login is provisioned for: ChatGPT-account
   logins use the config's codenamed ids (e.g. `gpt-5.6-sol`) and reject a bare
   `gpt-5.6`. When the user names a model ("review with 5.6-sol"), pass it here.
 - `CODEX_REVIEW_EFFORT` — reasoning effort, default `medium`. The probe and
   review stages accept only `medium` or `high`; any other value is rejected.
 - `CODEX_REVIEW_IDLE_SECONDS` / `CODEX_REVIEW_PROBE_IDLE_SECONDS` — idle timeout
   (seconds) for the review stage and the preflight probe, default `180` each.
-- `CODEX_REVIEW_ARTIFACT_ROOT` — directory for the recorded review transcript
-  used as outside-review evidence, default `<cwd>/.webpresso/reviews`.
+- `CODEX_REVIEW_ARTIFACT_ROOT` — directory for private runtime diagnostics,
+  default `<cwd>/.webpresso/reviews`. Never commit files from this directory or
+  cite them as approval evidence.
+
+Direct `wp review run` output is advisory. Only `wp review gate` may publish a
+bounded, target-bound review artifact and ledger event for blueprint promotion
+or completion.
 
 ## Modes
 
@@ -79,7 +85,11 @@ overrides (defaults shown). Set them inline, e.g.
 ```bash
 # CODEX_REVIEW_MODEL / CODEX_REVIEW_EFFORT are overridable — see ## Options for
 # the full option list, valid values, and how to pick a login-supported model.
-CODEX_REVIEW_MODEL=${CODEX_REVIEW_MODEL:-gpt-5.4}
+CODEX_REVIEW_MODEL=${CODEX_REVIEW_MODEL:-}
+CODEX_MODEL_ARGS=()
+if [ -n "$CODEX_REVIEW_MODEL" ]; then
+  CODEX_MODEL_ARGS=(--model "$CODEX_REVIEW_MODEL")
+fi
 CODEX_REVIEW_EFFORT=${CODEX_REVIEW_EFFORT:-medium}
 case "$CODEX_REVIEW_EFFORT" in
   medium|high) ;;
@@ -96,7 +106,7 @@ CODEX_REVIEW_CODE=0
 if wp review run \
   --provider codex \
   --prompt-file "$PROBE_FILE" \
-  --model "$CODEX_REVIEW_MODEL" \
+  "${CODEX_MODEL_ARGS[@]}" \
   --effort "$CODEX_REVIEW_EFFORT" \
   --stage probe \
   --expected-marker CODEX_OK \
@@ -105,7 +115,7 @@ if wp review run \
   wp review run \
     --provider codex \
     --prompt-file "$PROMPT_FILE" \
-    --model "$CODEX_REVIEW_MODEL" \
+    "${CODEX_MODEL_ARGS[@]}" \
     --effort "$CODEX_REVIEW_EFFORT" \
     --stage review \
     --artifact-root "$CODEX_REVIEW_ARTIFACT_ROOT" \

package/skills/grok/SKILL.md

@@ -0,0 +1,45 @@
+---
+name: grok
+description: "Grok Build outside-voice review/consult for non-Grok hosts."
+license: MIT
+---
+
+# Grok outside voice
+
+Use when a non-Grok host needs Grok Build to review a diff, challenge a plan, or answer a focused repo question. Keep it read-only by default. A provenance-backed Grok approval is first-class and satisfies the blueprint strong-reviewer gate alone, with the same rejection authority as Claude or Codex.
+
+## Auth / binary
+
+```bash
+if ! command -v grok >/dev/null 2>&1; then
+  echo "GROK_AUTH=missing: install Grok Build CLI (grok) and authenticate"
+  exit 1
+fi
+echo "GROK_AUTH=cli-present"
+```
+
+## Review via typed runtime
+
+```bash
+PROMPT_FILE=$(mktemp -t wp-grok-review.XXXXXX)
+trap 'rm -f "$PROMPT_FILE"' EXIT
+# write review prompt into PROMPT_FILE
+
+GROK_REVIEW_MODEL=${GROK_REVIEW_MODEL:-}
+GROK_MODEL_ARGS=()
+if [ -n "$GROK_REVIEW_MODEL" ]; then
+  GROK_MODEL_ARGS=(--model "$GROK_REVIEW_MODEL")
+fi
+GROK_REVIEW_EFFORT=${GROK_REVIEW_EFFORT:-medium}
+GROK_REVIEW_ARTIFACT_ROOT=${GROK_REVIEW_ARTIFACT_ROOT:-"$(pwd)/.webpresso/reviews"}
+
+wp review run \
+  --provider grok \
+  --prompt-file "$PROMPT_FILE" \
+  "${GROK_MODEL_ARGS[@]}" \
+  --effort "$GROK_REVIEW_EFFORT" \
+  --stage review \
+  --artifact-root "$GROK_REVIEW_ARTIFACT_ROOT"
+```
+
+Treat output as external advice until independently verified. For lifecycle authority, run it through `wp review gate --provider grok` so the verdict is bound to the reviewed subject and recorded in the v2 ledger.

package/skills/hooks-doctor/SKILL.md

@@ -55,13 +55,13 @@ Each check prints `[x]` (pass) or `[ ]` (fail) with a detail line:
 
 ## Failure Remediation
 
-| Check                                            | Likely Cause                                                   | Fix                                                            |
-| ------------------------------------------------ | -------------------------------------------------------------- | -------------------------------------------------------------- |
-| `pretool-guard` / `post-tool` / etc. — not found | `pnpm build` not run after install                             | `pnpm build`                                                   |
-| `pretool-guard` / etc. — not executable          | `chmod +x` not persisted                                       | Re-run `pnpm prepare` or `pnpm build` which runs `chmod-bins`  |
-| `plugin.json integrity` — missing                | Claude adapter manifest absent from `@webpresso/claude-plugin` | Re-run `wp setup --host claude` or reinstall the Claude plugin |
-| `MCP server liveness` — timeout                  | MCP server cold-start too slow                                 | Wait and retry, or run `wp hooks doctor --skip-mcp`            |
-| Any check — not found at `dist/esm/...`          | Build artifacts missing                                        | Run `pnpm build` in the webpresso repo                         |
+| Check                                            | Likely Cause                                                   | Fix                                                               |
+| ------------------------------------------------ | -------------------------------------------------------------- | ----------------------------------------------------------------- |
+| `pretool-guard` / `post-tool` / etc. — not found | `wp run build` not run after install                           | `wp run build`                                                    |
+| `pretool-guard` / etc. — not executable          | `chmod +x` not persisted                                       | Re-run `wp run prepare` or `wp run build` which runs `chmod-bins` |
+| `plugin.json integrity` — missing                | Claude adapter manifest absent from `@webpresso/claude-plugin` | Re-run `wp setup --host claude` or reinstall the Claude plugin    |
+| `MCP server liveness` — timeout                  | MCP server cold-start too slow                                 | Wait and retry, or run `wp hooks doctor --skip-mcp`               |
+| Any check — not found at `dist/esm/...`          | Build artifacts missing                                        | Run `wp run build` in the webpresso repo                          |
 
 After fixing, re-run `wp hooks doctor` to confirm.
 

package/skills/opencode-go/SKILL.md

@@ -57,6 +57,7 @@ The managed OpenCode plugin currently has no parent-linked native subagent compl
 ## Review command
 
 Use read-only prompts. Run from the repo directory; do NOT pass `--dir "$PWD"` because opencode already uses the current directory and the extra index can stall reviews.
+The artifact root contains private runtime diagnostics, not a recorded transcript or committable approval evidence. Direct `wp review run` output is advisory. Only `wp review gate` may publish a bounded, target-bound artifact and ledger event for blueprint promotion or completion.
 
 ```bash
 # Aggregate reviewer: honor the canonical reviewer policy (Kimi K2.7 Code → DeepSeek V4 Pro → DeepSeek V4 Flash first, then cross-family fallbacks).
@@ -106,4 +107,4 @@ wp review run \
 ```
 
 If `$MODEL` is empty, run `opencode models opencode-go` and choose a present reviewer.
-Treat a provider-wide usage limit as terminal for every OpenCode Go model in the current review; do not fall back to another family. Treat `true-idle`, `protocol-unsupported`, provider failure, abort, spawn failure, or artifact failure as unavailable review evidence. Never replace them with a static timeout, arbitrary byte-growth heartbeat, or buffered-output fallback.
+Treat a provider-wide usage limit as terminal for every OpenCode Go model in the current review; do not fall back to another family. Treat `true-idle`, `protocol-unsupported`, provider failure, abort, spawn failure, or artifact failure as an unavailable advisory result. Never replace them with a static timeout, arbitrary byte-growth heartbeat, or buffered-output fallback.

package/skills/verify/SKILL.md

@@ -176,12 +176,12 @@ Default policy:
 
 Reviewer preference:
 
-- When running from Codex, prefer **Claude** (fall back to one OpenCode Go reviewer if Claude is unavailable).
-- When running from Claude, prefer **Codex** (fall back to one OpenCode Go reviewer if Codex is unavailable).
-- When Codex/Claude outside voice is unavailable, use **one OpenCode Go
+- When running from Codex, prefer **Claude or Grok** (fall back to one OpenCode Go reviewer if both are unavailable).
+- When running from Claude, prefer **Codex or Grok** (fall back to one OpenCode Go reviewer if both are unavailable).
+- When Codex/Claude/Grok outside voice is unavailable, use **one OpenCode Go
   reviewer** if OpenCode is logged in.
 - Prefer the repo's outside-voice skills instead of hand-rolled commands:
-  `$agent-kit:claude`, `$agent-kit:codex`, and one OpenCode Go reviewer
+  `$agent-kit:claude`, `$agent-kit:codex`, `$agent-kit:grok`, and one OpenCode Go reviewer
   skills such as `$agent-kit:opencode-go`, `$agent-kit:qwen`,
   `$agent-kit:deepseek`, `$agent-kit:glm`, `$agent-kit:kimi`,
   `$agent-kit:minimax`, or `$agent-kit:mimo`.