feat(miner): add cross-repo --full-execution dry-run mode

apps/loopover-ui/src/lib/ams-env-reference.ts

@@ -91,6 +91,11 @@ export const AMS_ENV_REFERENCE_ROWS: MinerEnvReferenceRow[] = [
     firstReference: "lib/event-ledger.js",
     defaultValue: null,
   },
+  {
+    name: "LOOPOVER_MINER_FULL_EXECUTION_STUB",
+    firstReference: "lib/cross-repo-evaluation.js",
+    defaultValue: "",
+  },
   {
     name: "LOOPOVER_MINER_GOVERNOR_LEDGER_DB",
     firstReference: "lib/governor-ledger.js",
@@ -232,6 +237,7 @@ export const AMS_ENV_REFERENCE_MARKDOWN = [
   '| `LOOPOVER_MINER_DISCOVERY_SHARED_SECRET` | `lib/discovery-index-client.js` | `""` |',
   '| `LOOPOVER_MINER_DISCOVERY_TELEMETRY` | `lib/discovery-index-client.js` | `""` |',
   "| `LOOPOVER_MINER_EVENT_LEDGER_DB` | `lib/event-ledger.js` | (none) |",
+  '| `LOOPOVER_MINER_FULL_EXECUTION_STUB` | `lib/cross-repo-evaluation.js` | `""` |',
   "| `LOOPOVER_MINER_GOVERNOR_LEDGER_DB` | `lib/governor-ledger.js` | (none) |",
   "| `LOOPOVER_MINER_GOVERNOR_STATE_DB` | `lib/governor-state.js` | (none) |",
   '| `LOOPOVER_MINER_LOG_LEVEL` | `lib/logger.js` | `""` |',

packages/loopover-miner/DEPLOYMENT.md

@@ -197,6 +197,7 @@ and `LOOPOVER_MINER_CONFIG_DIR` are covered above under the fleet/state notes; t
 | `LOOPOVER_MINER_AMS_POLICY_PATH` | `lib/ams-policy.js` | Path to the operator's `.loopover-ams.yml` policy spec. |
 | `LOOPOVER_MINER_AMS_COLLECTOR_URL`, `LOOPOVER_MINER_AMS_COLLECTOR_TOKEN` | `lib/orb-export.js` | Endpoint + bearer for pushing fleet state to an AMS/Orb collector. |
 | `LOOPOVER_MINER_CHAT_ACTIONS` | `lib/chat-action-dispatch.js` | Truthy string enables the chat-action-dispatch flag (off by default). |
+| `LOOPOVER_MINER_FULL_EXECUTION_STUB` | `lib/cross-repo-evaluation.js` | Truthy string enables a synthetic one-file handoff for local full-execution demos (off by default; never invents a live coding-agent run or forge PR). |
 | `LOOPOVER_MINER_LEDGER_RETENTION_DAYS`, `LOOPOVER_MINER_LEDGER_RETENTION_MAX_ROWS` | `lib/store-maintenance.js` | Opt-in ledger retention window / row cap. |
 | `LOOPOVER_MINER_LOG_LEVEL` | `lib/logger.js` | Log verbosity override. |
 | `LOOPOVER_MINER_NO_UPDATE_CHECK` | `lib/update-check.js` | Truthy string disables the background update check. |

packages/loopover-miner/benchmarks/cross-repo/manifest.json

@@ -1,5 +1,5 @@
 {
-  "description": "Cross-repo evaluation benchmark set (#4788). Diverse public repos exercised by `node packages/loopover-miner/scripts/cross-repo-evaluation.mjs` after cloning into LOOPOVER_MINER_REPO_CLONE_DIR.",
+  "description": "Cross-repo evaluation benchmark set (#4788 readiness / #7634 full-execution). Diverse public repos exercised by `node packages/loopover-miner/scripts/cross-repo-evaluation.mjs` after cloning into LOOPOVER_MINER_REPO_CLONE_DIR. Entries with fullExecution=true are the default --full-execution subset (dry-run: no forge PR writes).",
   "repos": [
     {
       "repoFullName": "sindresorhus/is",
@@ -8,7 +8,8 @@
     {
       "repoFullName": "expressjs/express",
       "stackHint": "nodejs",
-      "requireTestCommand": true
+      "requireTestCommand": true,
+      "fullExecution": true
     },
     {
       "repoFullName": "pallets/flask",
@@ -28,7 +29,8 @@
     {
       "repoFullName": "lodash/lodash",
       "stackHint": "nodejs",
-      "requireTestCommand": true
+      "requireTestCommand": true,
+      "fullExecution": true
     }
   ]
 }

packages/loopover-miner/docs/cross-repo-evaluation.md

@@ -20,6 +20,10 @@ Each benchmark repo receives a **pass/fail** line. Failures are categorized:
 | `loopover_assumption` | Agent instructions leak loopover-specific CI assumptions |
 | `clone_setup` | The repo has not been cloned to the expected cache path |
 | `other` | Unexpected errors |
+| `plan_formed_compile_failed` | Full-execution: local build/compile failed after a coding handoff (#7634) |
+| `compiled_tests_failed` | Full-execution: build ok but the repo's own test command failed (#7634) |
+| `tests_passed_noop_diff` | Full-execution: build+tests ok but the coding attempt changed no files (#7634) |
+| `execution_abandon` | Full-execution: coding attempt abandoned before handoff (#7634) |
 
 The run also reports whether a **strict majority** of repos passed and how many succeeded **without** a per-target
 `.loopover-miner.yml` (the default goal spec is acceptable).
@@ -33,6 +37,8 @@ bare `"owner/repo"` string or an object:
 - **`stackHint`** — documentation only (not used by the evaluator)
 - **`requireTestCommand`** — when `true`, stack detection must infer a test command or the repo fails with
   `execution_gap`
+- **`fullExecution`** — when `true`, include this repo in the default `--full-execution` subset (#7634). At least two
+  shipped entries are tagged; if fewer than two are tagged, the harness falls back to all `requireTestCommand` entries.
 
 Malformed manifest fields degrade to documented defaults with warnings (same tolerant-parser convention as the
 fleet run-manifest).
@@ -60,19 +66,40 @@ fleet run-manifest).
    - `--repo owner/repo` — evaluate a single manifest entry
    - `--manifest path/to/manifest.json` — alternate benchmark set (e.g. a fixture manifest in tests)
    - `--require-majority` — exit `1` unless a strict majority of repos pass (for CI-style gating)
+   - `--full-execution` — after readiness, run local discover→plan→code→build→test (**no forge PR writes**, #7634)
+
+## Full-execution mode (#7634)
+
+`--full-execution` extends readiness with a **local-only** code→build→test loop against the tagged benchmark subset
+(or `--repo` for a single entry). It never opens a GitHub/GitLab PR and never calls forge write APIs against the
+benchmark repos.
+
+1. Run the same readiness checks as the default mode
+2. Invoke an injectable `runCodingAttempt` seam (discover→plan→code) that must return `handoff` + `changedFiles`
+3. Run the detected `buildCommand` (when present) and `testCommand` locally via `runShellCommand`
+4. Fail with execution-specific categories when compile fails, tests fail, the diff is empty, or the attempt abandons
+
+**Default coding seam:** without an injected `runCodingAttempt`, the harness abandons with `execution_abandon`
+unless `LOOPOVER_MINER_FULL_EXECUTION_STUB=1` is set (synthetic one-file handoff for local demos). Real agent wiring
+stays opt-in so CI and unit tests stay offline and forge-write-free.
 
 ## Library API
 
 Pure functions live in [`lib/cross-repo-evaluation.js`](../lib/cross-repo-evaluation.js):
 
 - `parseCrossRepoEvaluationManifest(content)`
 - `evaluateRepoReadiness(entry, options)` — inject `existsSync`, `detectRepoStack`, etc. for unit tests
-- `runCrossRepoEvaluation(parsed, options)`
+- `evaluateRepoFullExecution(entry, options)` — readiness + local code→build→test; inject `runCodingAttempt` /
+  `runShellCommand` (#7634)
+- `runCrossRepoEvaluation(parsed, options)` — async; pass `fullExecution: true` for the execution path
 - `summarizeCrossRepoEvaluation(results)`
 - `formatCrossRepoEvaluationReport(results, summary)`
 
 ## Wiring
 
-This harness is **readiness-only**: it does not run the coding agent, open PRs, or call forge APIs. A green report
-means the miner’s repo-agnostic stack-detection and coding-task-spec path is prepared for the benchmark repo; a live
+**Readiness mode** does not run the coding agent, open PRs, or call forge APIs. A green report means the miner’s
+repo-agnostic stack-detection and coding-task-spec path is prepared for the benchmark repo.
+
+**Full-execution mode** still does not open PRs or call forge write APIs. A green report means a local coding
+handoff produced a non-empty diff that survived the target repo’s own build and test commands. A live production
 attempt still needs credentials, governor policy, and queue state as documented in [`DEPLOYMENT.md`](../DEPLOYMENT.md).

packages/loopover-miner/docs/env-reference.md

@@ -21,6 +21,7 @@ Generated by `npm run miner:env-reference`. Do not edit manually.
 | `LOOPOVER_MINER_DISCOVERY_SHARED_SECRET` | `lib/discovery-index-client.js` | `""` |
 | `LOOPOVER_MINER_DISCOVERY_TELEMETRY` | `lib/discovery-index-client.js` | `""` |
 | `LOOPOVER_MINER_EVENT_LEDGER_DB` | `lib/event-ledger.js` | (none) |
+| `LOOPOVER_MINER_FULL_EXECUTION_STUB` | `lib/cross-repo-evaluation.js` | `""` |
 | `LOOPOVER_MINER_GOVERNOR_LEDGER_DB` | `lib/governor-ledger.js` | (none) |
 | `LOOPOVER_MINER_GOVERNOR_STATE_DB` | `lib/governor-state.js` | (none) |
 | `LOOPOVER_MINER_LOG_LEVEL` | `lib/logger.js` | `""` |

packages/loopover-miner/lib/cross-repo-evaluation.d.ts

@@ -1,11 +1,19 @@
 import type { RepoStackResult } from "./stack-detection.js";
-/** Failure taxonomy surfaced in per-repo reports (#4788). */
+/** Failure taxonomy surfaced in per-repo reports (#4788 readiness + #7634 full-execution). */
 export declare const CROSS_REPO_FAILURE_CATEGORY: Readonly<{
     STACK_DETECTION: "stack_detection_gap";
     EXECUTION: "execution_gap";
     GITTENSOR_ASSUMPTION: "loopover_assumption";
     CLONE_SETUP: "clone_setup";
     OTHER: "other";
+    /** Plan/spec formed but local build/compile command failed (#7634). */
+    COMPILE_FAILED: "plan_formed_compile_failed";
+    /** Build succeeded but the repo's own test suite failed (#7634). */
+    TESTS_FAILED: "compiled_tests_failed";
+    /** Tests passed but the coding attempt produced no file changes (#7634). */
+    NOOP_DIFF: "tests_passed_noop_diff";
+    /** Coding attempt abandoned before a handoff (agent unconfigured / refused) (#7634). */
+    EXECUTION_ABANDON: "execution_abandon";
 }>;
 /** Instruction substrings that indicate a POSITIVE loopover/LoopOver CI assumption leaked into the agent prompt.
  *  Lines that explicitly tell the agent *not* to assume these are filtered out before scanning. */
@@ -20,6 +28,8 @@ export type CrossRepoEvaluationManifestRepo = {
     repoFullName: string;
     stackHint?: string;
     requireTestCommand?: boolean;
+    /** When true, include this repo in `--full-execution` runs (#7634). */
+    fullExecution?: boolean;
     fixturePath?: string;
 };
 export type ParsedCrossRepoEvaluationManifest = {
@@ -67,6 +77,36 @@ type EvaluateRepoReadinessOptions = {
         instructions?: string;
     };
 };
+/** Local coding-attempt outcome for `--full-execution` (#7634). Never opens a forge PR. */
+export type CrossRepoCodingAttemptResult = {
+    outcome: "handoff" | "abandon";
+    changedFiles: string[];
+    reason?: string;
+};
+export type CrossRepoShellCommandResult = {
+    ok: boolean;
+    exitCode: number;
+    stdout?: string;
+    stderr?: string;
+};
+export type EvaluateRepoFullExecutionOptions = EvaluateRepoReadinessOptions & {
+    /**
+     * Run the local discover→plan→code step and return changed files. MUST NOT open PRs or call forge
+     * write APIs (#7634). Injected in unit tests; default is a no-network stub that abandons unless
+     * `LOOPOVER_MINER_FULL_EXECUTION_STUB=1` (synthetic handoff for local dry demos).
+     */
+    runCodingAttempt?: (input: {
+        repoFullName: string;
+        repoPath: string;
+        stack: RepoStackResult;
+        instructions: string;
+    }) => CrossRepoCodingAttemptResult | Promise<CrossRepoCodingAttemptResult>;
+    /** Run a local shell command (build/test). Injected in unit tests. */
+    runShellCommand?: (input: {
+        command: string;
+        cwd: string;
+    }) => CrossRepoShellCommandResult | Promise<CrossRepoShellCommandResult>;
+};
 /** Canonical `owner/repo` with exactly one slash and safe segments; anything else → null. */
 export declare function normalizeCrossRepoFullName(value: unknown): string | null;
 /**
@@ -86,12 +126,39 @@ export declare function scanPositiveLoopoverAssumptions(text: string): Array<{
  * Evaluate one benchmark repo's miner readiness without running a live coding agent (#4788).
  */
 export declare function evaluateRepoReadiness(entry: CrossRepoEvaluationManifestRepo, options?: EvaluateRepoReadinessOptions): CrossRepoEvaluationResult;
+/**
+ * Default coding-attempt seam for `--full-execution` (#7634). Never opens a PR.
+ * - `LOOPOVER_MINER_FULL_EXECUTION_STUB=1` → synthetic handoff with one changed path (local demo only).
+ * - Otherwise abandons with a clear reason (real agent wiring stays opt-in via `runCodingAttempt`).
+ */
+export declare function defaultRunCodingAttempt(input: {
+    repoFullName: string;
+    repoPath: string;
+    stack: RepoStackResult;
+    instructions: string;
+    env?: NodeJS.ProcessEnv;
+}): CrossRepoCodingAttemptResult;
+/**
+ * Default local shell runner for build/test commands (#7634). Sync spawn; no network.
+ */
+export declare function defaultRunShellCommand(input: {
+    command: string;
+    cwd: string;
+}): CrossRepoShellCommandResult;
+/**
+ * Full-execution evaluation (#7634): readiness first, then local code→build→test with no forge writes.
+ */
+export declare function evaluateRepoFullExecution(entry: CrossRepoEvaluationManifestRepo, options?: EvaluateRepoFullExecutionOptions): Promise<CrossRepoEvaluationResult>;
 /**
  * Run the harness across every repo in a parsed manifest (#4788).
+ * Pass `fullExecution: true` (#7634) to run local code→build→test (still no forge writes).
  */
 export declare function runCrossRepoEvaluation(parsed: ParsedCrossRepoEvaluationManifest, options?: {
     repoFilter?: string;
-} & EvaluateRepoReadinessOptions): CrossRepoEvaluationResult[];
+    fullExecution?: boolean;
+    /** Minimum repos to include in full-execution when filtering by `fullExecution` / requireTestCommand. */
+    fullExecutionMinRepos?: number;
+} & EvaluateRepoFullExecutionOptions): Promise<CrossRepoEvaluationResult[]>;
 /**
  * Reduce per-repo results to pass/fail counts and whether a strict majority passed (#4788).
  */