@amsterdamdatalabs/enact-extensions 0.1.25 → 0.1.27

package/extensions/enact-evolve/skills/mined-investigate-loop/SKILL.md

@@ -0,0 +1,68 @@
+---
+name: mined-investigate-loop
+description: >-
+  Auto-mined procedure (enact-evolve P4 procedural tier). Before editing
+  unfamiliar code, run an investigate loop: read a file, grep to locate the
+  related symbol/usage, then read the located target — read → grep → read —
+  before making changes. Mined from 251 real Claude+Codex sessions; sessions
+  that ran this loop succeeded at 58.3% vs a 43.0% baseline (+15pp). Use when
+  navigating unfamiliar code, locating a definition/usage, or scoping a change
+  before editing. Trigger keywords: investigate, locate, find usages, where is,
+  scope a change, read before edit, navigate code.
+---
+
+# mined-investigate-loop (read → grep → read)
+
+> **Provenance — machine-mined, pending human graduation (Track B G2).** This
+> skill was produced by the enact-evolve procedural tier's execution-mining →
+> graduation path over real session history. It is under review; do not treat it
+> as hand-authored guidance until merged.
+
+## The procedure
+
+When you are about to change code you don't already hold in context, **investigate
+before you act**:
+
+1. **read** — open the entry file you were handed (the symbol, the failing test,
+   the reported file).
+2. **grep** — search the repo for the related symbol / call site / definition you
+   now know to look for (don't guess the path; locate it).
+3. **read** — open the located target and confirm the real shape before editing.
+
+Only then edit. The loop may repeat (read → grep → read → grep → …) until the
+change is fully scoped.
+
+## When to use
+
+- Navigating unfamiliar code or a new package.
+- Locating a definition, its usages, or the blast radius of a change.
+- Scoping an edit before making it (especially across files).
+
+## When NOT to use
+
+- You already have the target file and its dependencies in context — skip
+  straight to the edit.
+- Pure debugging-thrash (grep → stream-edit → grep with no reads) is an
+  **anti-pattern** the miner explicitly rejected (see Evidence) — investigate by
+  reading, don't blind-edit-and-re-grep.
+
+## Evidence (why this graduated)
+
+Mined by `enact-evolve/eval/mine-procedures.mjs` over real session transcripts
+(the existing Claude + Codex adapters → `SessionEvidence.tool_timeline`):
+
+| field | value |
+|---|---|
+| corpus | 251 sessions (205 Claude + 46 Codex) |
+| baseline zero-failure rate | 0.430 |
+| this procedure (read → grep → read) | **60 sessions, 35 success → 0.583** |
+| lift over baseline | **+0.153 (×1.36)** |
+| graduation policy | rate ≥ 1.25× baseline, support ≥ 8, success ≥ 4 |
+
+Reward signal: a session "succeeds" when it records zero failed tool/hook/watchdog
+events (`SessionEvidence.failures` empty). Graduation is **lift-based** because the
+absolute zero-failure base rate is low; frequency alone is rejected — the
+highest-frequency pattern `rg → sed → rg` (debugging thrash) scored 2/13 (0.15)
+and did **not** graduate.
+
+Full artifact: `enact-evolve/eval/P4-mined-procedures.json`.

package/extensions/enact-factory/skills/autonomous-runner/SKILL.md

@@ -215,14 +215,19 @@ or return an empty session:
 
 | Grader provider | Paseo flag | Notes |
 |---|---|---|
-| `codex` | `--mode auto` | `--mode bypass` is rejected by codex |
+| `codex` | `--mode auto-review` | Routes the grader's eligible `on-request` approvals through codex's auto-reviewer subagent instead of BLOCKING — without it a grader can stall on an approval (e.g. a cargo/network fetch during build verification). `--mode bypass` is rejected by codex |
 | `cursor` | default / agent | `--mode bypass` is a claude mode, NOT cursor's — previously documented wrong; cursor uses default or agent mode |
 | `kimi` | `--mode auto` | Same as codex |
 | `opencode` (minimax M3) | default | No special mode flag needed |
 
+Codex mode enum (authoritative): `read-only` · `auto` · `auto-review` · `full-access`. The grader
+stays read-only BY PROMPT; `auto-review` only governs approval handling and composes with
+`--output-schema` verdict capture. Flip a running grader with `paseo agent mode <id> auto-review`.
+
 **Verdict capture is STRUCTURED, not log-scraped.** Dispatch the grader with:
-`paseo run --output-schema <verdict.json> --json --label continuation=<cid>`
-— stdout is the validated `{verdict, evidence}` object. Recover the actual grader model and cost
+`paseo run --provider <graderModel> --mode auto-review --output-schema <verdict.json> --json --label continuation=<cid>`
+— stdout is the validated `{verdict, evidence}` object. On the codex provider, `--mode auto-review`
+keeps on-request approvals from blocking the grader (read-only stays enforced by prompt). Recover the actual grader model and cost
 via `paseo ls -a --label continuation=<cid> --json` then `paseo inspect <id> --json`. Watch live
 with `paseo attach <id>` (recover the id by label first).
 See design doc: `enact-loop/docs/design/2026-06-22-paseo-verdict-capture.md`.
@@ -243,7 +248,8 @@ lists in `FACTORY_DRIVE_DECISION_TREE.md` §4), then hand to
 - `loop_start({ contract })` — the boulder activates; you cannot stop until
   every required stage passes and every judgment stage has an independent GO.
 - Mechanical stages → `loop_grade({ stageId })` — engine runs stage command in loop root.
-- Judgment stages → `loop_grader_dispatch` → spawn grader on a **different model** →
+- Judgment stages → `loop_grader_dispatch` → spawn grader on a **different model**
+  (Paseo codex provider: `--mode auto-review` so on-request approvals don't block) →
   driver relays `loop_grader_verdict` (`captureSource: "relayed"`) or grader calls directly.
 - **No `git stash` during an active loop** — invalidates worktree fingerprints.
 - On stall: escalate via `committee` (root-cause) or `advisor` (judgment call);

package/extensions/enact-factory/skills/drive-loop/SKILL.md

@@ -161,7 +161,7 @@ For each judgment stage:
 > the factory MCP tool `factory_paseo_grader_run` runs the grader with
 > `paseo run --output-schema --json --wait-timeout` so **stdout is the validated `{verdict,
 > evidence}` object** — no `paseo logs | grep`. It bakes in the per-provider `--mode`
-> (codex/kimi `auto`, cursor/opencode default) and recovers the actual resolved model via
+> (codex `auto-review` so on-request approvals don't block, kimi `auto`, cursor/opencode default) and recovers the actual resolved model via
 > `paseo ls --label` / `inspect`, which you relay as `graderModelActual` on `loop_grader_verdict`.
 > Full design: `enact-loop/docs/design/2026-06-22-paseo-verdict-capture.md`.
 
@@ -250,11 +250,15 @@ OpenAI lineage); grade sonnet work with gpt-5.5 / opus / minimax / kimi.
 
 | Provider | Paseo flag | Notes |
 |---|---|---|
-| `codex` | `--mode auto` | `--mode bypass` is rejected by codex |
+| `codex` | `--mode auto-review` | Routes the grader's eligible `on-request` approvals through codex's auto-reviewer subagent instead of BLOCKING — without it a grader can stall on an approval (e.g. a cargo/network fetch during build verification). `--mode bypass` is rejected by codex |
 | `cursor` | default / agent | `--mode bypass` is a claude mode, NOT cursor's — previously documented wrong |
 | `kimi` | `--mode auto` | Same as codex |
 | `opencode` (minimax M3) | default | No special mode flag needed |
 
+Codex mode enum (authoritative): `read-only` · `auto` · `auto-review` · `full-access`. The grader
+stays read-only BY PROMPT; `auto-review` only governs approval handling and composes with
+`--output-schema` verdict capture. Flip a running grader with `paseo agent mode <id> auto-review`.
+
 #### Sonnet-executor caveat
 
 `cursor` (`cursor/claude-sonnet-4-6`) is NOT model-distinct from a sonnet executor.
@@ -264,9 +268,10 @@ escalate to opus 4.8 / gpt-5.5 for high-risk judgment. Reserve `cursor` for opus
 #### Verdict capture — structured `--output-schema`, not log-scraped
 
 Dispatch the grader with:
-`paseo run --output-schema <verdict.json> --json --label continuation=<cid>`
+`paseo run --provider <graderModel> --mode auto-review --output-schema <verdict.json> --json --label continuation=<cid>`
 — stdout is the validated `{verdict, evidence}` object (verified on codex / cursor / opencode / kimi;
-coexists with tool use). Recover actual grader model + cost via
+coexists with tool use). On the codex provider, `--mode auto-review` keeps on-request approvals from
+blocking the grader (read-only stays enforced by prompt). Recover actual grader model + cost via
 `paseo ls -a --label continuation=<cid> --json` + `paseo inspect <id> --json`. Watch live by
 recovering the id by label then `paseo attach <id>`.
 Design doc: `enact-loop/docs/design/2026-06-22-paseo-verdict-capture.md`.

package/extensions/enact-factory/skills/drive-loop/references/contract-schema.md

@@ -37,7 +37,7 @@ disclosure supplement to `SKILL.md`.
 | `role` | `"architect" \| "critic" \| "code-reviewer" \| "verifier"` | yes | Determines the subagent type spawned. |
 | `model` | string | recommended | Model ID for the grader. **Must differ from the executor model.** Cross-vendor preferred. Omit only if the harness selects automatically. |
 | `harness` | `"paseo" \| "subagent"` | optional | How to dispatch. Default: `subagent`. Use `paseo` when the grader needs multi-step tool access. |
-| `dispatchTool` | string | optional | Factory tool the driver uses to run a Paseo grader and capture a structured verdict — `"factory_paseo_grader_run"` (runs `paseo run --output-schema --json`, returns `{verdict, evidence}` + `graderModelActual`). Replaces `paseo logs` scraping. |
+| `dispatchTool` | string | optional | Factory tool the driver uses to run a Paseo grader and capture a structured verdict — `"factory_paseo_grader_run"` (runs `paseo run --provider <graderModel> --mode auto-review --output-schema --json`, returns `{verdict, evidence}` + `graderModelActual`). On the codex provider `--mode auto-review` routes on-request approvals through the auto-reviewer subagent so the grader doesn't BLOCK (read-only stays enforced by prompt). Replaces `paseo logs` scraping. |
 | `rounds` | integer | optional | Number of independent grading rounds before aggregating. Default: `1`. |
 | `timeoutMs` | integer | optional | Grader session timeout in ms (default 30 min). Timeout → `grader-unavailable` blocker. |
 

package/extensions/enact-factory/skills/handoff/SKILL.md

@@ -35,7 +35,10 @@ the factory value-add the generic delegate pattern lacks.
   repos, change scope, or split the item. Out-of-scope discoveries → new
   WorkItem (the receiver follows the same rule).
 - **The receiver still drives the loop** (its own contract + cross-vendor
-  grader). Handoff delegates the work, not the gates.
+  grader). Handoff delegates the work, not the gates. When the receiver
+  dispatches a grader via Paseo on the codex provider, it passes
+  `--mode auto-review` so the grader's on-request approvals don't block (the
+  grader stays read-only by prompt); see `drive-loop`.
 
 ## Sequence
 

package/extensions/enact-loop/skills/enact-loop/SKILL.md

@@ -72,7 +72,8 @@ identity.
 
 Judgment stages must be graded by a **different model** than the executor (model-distinct, not
 vendor/surface-distinct). The loop rejects a verdict whose grader model, session, or agent equals
-the executor's. Graders are dispatched out-of-session via Paseo `--provider <model>` for
+the executor's. Graders are dispatched out-of-session via Paseo `--provider <model>` (use
+`--mode auto-review` so on-request approvals don't block; see Verdict capture) for
 cross-surface grading; in-surface graders just need a supported model that differs from the
 executor's.
 
@@ -143,6 +144,7 @@ loop_grader_dispatch  stageId=<id>  [model=<grader-model>]
   advancing through `fallbackModels` on re-dispatch after a spawn failure
 - requires the resolved model to differ from the executor's (`E_NO_DIVERSITY` otherwise)
 - launches the grader agent (read-only, same cwd) via Paseo `--provider <graderModel>`
+  (use `--mode auto-review` so on-request approvals don't block; see Verdict capture)
 - returns the `continuationId` you must pass to step 2
 
 **Step 2 — Record the grader's verdict** (driver-relayed is the default):
@@ -228,9 +230,15 @@ They MUST NOT call `loop_*` management tools (complete, pause, resume, retarget,
 or `factory_*` tools. The runtime rejects these calls from grader sessions.
 
 **Verdict capture is STRUCTURED, not log-scraped.** Dispatch with:
-`paseo run --output-schema <verdict.json> --json --label continuation=<cid>`
+`paseo run --provider <graderModel> --mode auto-review --output-schema <verdict.json> --json --label continuation=<cid>`
 — stdout is the validated `{verdict, evidence}` object (verified on codex / cursor / opencode / kimi;
-coexists with tool use). Recover actual grader model + cost via
+coexists with tool use). With the codex provider, `--mode auto-review` routes the grader's eligible
+`on-request` approvals through codex's auto-reviewer subagent instead of BLOCKING — without it a
+grader can stall waiting on an approval (e.g. a cargo/network fetch during build verification). The
+grader stays read-only BY PROMPT; auto-review only governs approval handling and composes with the
+`--output-schema` structured verdict capture. Codex mode enum (authoritative): `read-only` · `auto`
+· `auto-review` · `full-access`. Flip a running grader with `paseo agent mode <id> auto-review`.
+Recover actual grader model + cost via
 `paseo ls -a --label continuation=<cid> --json` + `paseo inspect <id> --json`.
 Watch live: recover the id by label then `paseo attach <id>`.
 Fallback only: if structured capture is unavailable, embed evidence inline in the `evidence` field.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@amsterdamdatalabs/enact-extensions",
-  "version": "0.1.25",
+  "version": "0.1.27",
   "description": "Create and validate Enact multi-platform plugin manifests",
   "license": "UNLICENSED",
   "type": "module",