@kylecheng3146/agent-ops 0.1.15 → 0.1.17

package/README.md

@@ -287,14 +287,16 @@ task criteria and require current verification evidence. The detailed report
 is displayed; a minimal source-fingerprint attestation is persisted after PASS.
 
 Each attempt uses a fresh temporary cwd, a small allowlisted environment, and a
-target-native read-only/context-isolation mode. Currently only Claude safe mode
-meets the full isolation contract; configured Codex and Agy entries return
-`capability-unavailable` rather than run with a weaker boundary. `opencode` is
-not a review target.
-
-The first target that actually runs produces the verdict. A `FAIL` is final:
-the chain never retries elsewhere after a real verdict. `--yes` is still
-required for every run, since each run spends another provider's quota.
+target-native read-only mode. Claude additionally uses its complete safe mode;
+Codex ignores user config and persistence, while Agy runs in sandboxed plan
+mode against a disposable repository clone. Codex and Agy preserve their
+existing login environment, so their context isolation is intentionally weaker
+than Claude's. `opencode` is not a review target.
+
+The first valid `PASS` or `FAIL` is final. Attempts that produce no verdict —
+including login failures and unparseable output — advance to the next target
+and remain visible in the result's `attempts`. `--yes` is still required for
+every run, since each run spends another provider's quota.
 
 Authentication is diagnosed, never guessed:
 

package/dist/packages/cli/src/args.js

@@ -63,6 +63,7 @@ export function parseArgs(argv) {
     let harness;
     const hookTargets = [];
     let taskId;
+    let parentTaskId;
     let targetVersion;
     let title;
     let sessionId;
@@ -146,6 +147,14 @@ export function parseArgs(argv) {
                 index += 1;
                 break;
             }
+            case "--parent": {
+                if (parentTaskId !== undefined) {
+                    duplicate(token);
+                }
+                parentTaskId = readOptionValue(argv, index, token);
+                index += 1;
+                break;
+            }
             case "--target-version": {
                 if (targetVersion !== undefined) {
                     duplicate(token);
@@ -276,6 +285,7 @@ export function parseArgs(argv) {
             hookTargets.length > 0 ||
             profiles.length > 0 ||
             taskId !== undefined ||
+            parentTaskId !== undefined ||
             targetVersion !== undefined ||
             title !== undefined ||
             criteria.length > 0 ||
@@ -310,6 +320,9 @@ export function parseArgs(argv) {
         (hasTaskTargetOptions || hasTaskMutationOptions)) {
         throw new CliArgumentError("CLI_OPTION_NOT_ALLOWED", "Task target options may be used only with task or verify.");
     }
+    if (command !== "task" && parentTaskId !== undefined) {
+        throw new CliArgumentError("CLI_OPTION_NOT_ALLOWED", "--parent may be used only with task create or task status.");
+    }
     if (command !== "update" && targetVersion !== undefined) {
         throw new CliArgumentError("CLI_OPTION_NOT_ALLOWED", "--target-version may be used only with update.");
     }
@@ -342,20 +355,26 @@ export function parseArgs(argv) {
                 (title !== undefined ||
                     criteria.length > 0 ||
                     evidence.length > 0 ||
+                    // --parent filters the listing, so it cannot name a single target.
+                    (parentTaskId !== undefined &&
+                        (taskId !== undefined || sessionId !== undefined)) ||
                     (taskId !== undefined && sessionId !== undefined))) ||
             (action === "attach" &&
                 (title !== undefined ||
                     criteria.length > 0 ||
-                    evidence.length > 0)) ||
+                    evidence.length > 0 ||
+                    parentTaskId !== undefined)) ||
             (action === "complete" &&
                 (title !== undefined ||
                     criteria.length > 0 ||
-                    sessionId !== undefined)) ||
+                    sessionId !== undefined ||
+                    parentTaskId !== undefined)) ||
             ((action === "archive" || action === "export") &&
                 (title !== undefined ||
                     criteria.length > 0 ||
                     evidence.length > 0 ||
-                    sessionId !== undefined));
+                    sessionId !== undefined ||
+                    parentTaskId !== undefined));
         if (taskOptionInvalid) {
             throw new CliArgumentError("CLI_OPTION_NOT_ALLOWED", `Unsupported option for task ${action}.`);
         }
@@ -392,6 +411,7 @@ export function parseArgs(argv) {
         ...(hookTargets.length === 0 ? {} : { hookTargets }),
         profiles,
         ...(taskId === undefined ? {} : { taskId }),
+        ...(parentTaskId === undefined ? {} : { parentTaskId }),
         ...(targetVersion === undefined ? {} : { targetVersion }),
         ...(title === undefined ? {} : { title }),
         ...(reviewTargets.length === 0 ? {} : { reviewTargets }),

package/dist/packages/cli/src/cli.js

@@ -38,6 +38,8 @@ Options:
   --check-auth                        Doctor only: probe each review target's
                                       authentication with one real call
   --task <id>
+  --parent <task-id>                  Task create: record a subtask of this
+                                      task; task status: list its subtasks
   --target-version <version>          Update target version (offline-capable)
   --title <text>
   --criterion <json>                    Repeatable

package/dist/packages/cli/src/commands/review.js

@@ -217,6 +217,7 @@ function sourceChangedResult(result) {
         ...(result.independence === undefined
             ? {}
             : { independence: result.independence }),
+        ...(result.attempts === undefined ? {} : { attempts: result.attempts }),
         ...(result.verification === undefined
             ? {}
             : { verification: result.verification })
@@ -357,7 +358,10 @@ export async function runReviewCommand(options) {
                 });
             }
             try {
-                await assertSafeSupportingPaths(options.root, result.report.supportingFilesInspected);
+                await assertSafeSupportingPaths(options.root, [
+                    ...result.report.supportingFilesInspected,
+                    ...(result.adversarial?.report.supportingFilesInspected ?? [])
+                ]);
             }
             catch (error) {
                 if (scopeReason(error) === "unsafe-review-path") {

package/dist/packages/cli/src/commands/task.js

@@ -92,16 +92,22 @@ export async function runTaskCommand(options) {
                 criteria: (options.args.criteria ?? []).map(parseCriterion),
                 ...(options.policyConfigHash === undefined
                     ? {}
-                    : { policyConfigHash: options.policyConfigHash })
+                    : { policyConfigHash: options.policyConfigHash }),
+                ...(options.args.parentTaskId === undefined
+                    ? {}
+                    : { parentTaskId: options.args.parentTaskId })
             });
             return taskEnvelope(action, "TASK_CREATED", `Created task ${record.task.id}.`, record);
         }
         if (action === "status") {
             if (options.args.taskId === undefined && sessionId === undefined) {
-                const records = await options.service.list();
+                const parentTaskId = options.args.parentTaskId;
+                const records = await options.service.list(parentTaskId === undefined ? {} : { parentTaskId });
                 return okEnvelope("TASK_LISTED", {
                     action,
-                    message: `Listed ${records.length} task(s).`,
+                    message: parentTaskId === undefined
+                        ? `Listed ${records.length} task(s).`
+                        : `Listed ${records.length} subtask(s) of ${parentTaskId}.`,
                     records,
                     text: renderTaskList(records)
                 });

package/dist/packages/cli/src/wizard.js

@@ -1,18 +1,18 @@
 import { CliArgumentError } from "./args.js";
-import { DEFAULT_REVIEW_TARGETS } from "../../../runtime/src/review/roles.js";
+import { DEFAULT_REVIEW_TARGETS, REVIEW_TARGET_ORDER } from "../../../runtime/src/review/roles.js";
 import { HARNESS_IDS, resolveHarnessSelection } from "../../../runtime/src/install/harness.js";
 import { selectOption, selectOptions } from "./ui.js";
 const SCOPES = new Set(["project", "user"]);
 const PROFILES = new Set(["advisory", "core", "guardrails", "loop"]);
 const DEFAULT_HARNESS = [];
-const REVIEW_TARGET_SET = new Set(DEFAULT_REVIEW_TARGETS);
+const REVIEW_TARGET_SET = new Set(REVIEW_TARGET_ORDER);
 const REVIEW_TARGET_CHOICES = DEFAULT_REVIEW_TARGETS.map((id) => ({
     label: id,
     value: id,
     description: id === "codex"
-        ? "Retained for future isolation support; currently not auto-run."
+        ? "Runs read-only from its own credential home."
         : id === "agy"
-            ? "Retained for future isolation support; currently not auto-run."
+            ? "Runs in sandboxed plan mode against a disposable clone."
             : "Runs in fresh safe mode with context isolation."
 }));
 function selectReviewTargets(raw) {
@@ -26,7 +26,7 @@ function selectReviewTargets(raw) {
         }
     }
     // Declared order wins: the chain order is the option list, not click order.
-    return DEFAULT_REVIEW_TARGETS.filter((target) => values.includes(target));
+    return REVIEW_TARGET_ORDER.filter((target) => values.includes(target));
 }
 function affirmative(raw) {
     return /^(y|yes)$/i.test(raw.trim());

package/dist/runtime/src/install/doctor.js

@@ -370,21 +370,35 @@ async function checkReviewTargets(config, probe, checkAuth) {
         return check("review-targets", "PASS", `External review targets: ${targets.join(", ")}. ` +
             "Login state unverified; run: agent-ops doctor --check-auth");
     }
+    // Every target is probed before reporting: an ineligible entry is only
+    // DEGRADED, so returning at the first one would hide a genuinely broken
+    // eligible target behind it and make the verdict depend on ordering.
+    const failures = [];
+    const degraded = [];
     for (const target of targets) {
         const result = await probe(target, checkAuth);
         if (result === "missing-executable") {
-            return check("review-targets", "FAIL", `${target} not found.`, undefined, `Install ${target}, or remove "${target}" from reviewRoles[].targets.`);
+            failures.push(check("review-targets", "FAIL", `${target} not found.`, undefined, `Install ${target}, or remove "${target}" from reviewRoles[].targets.`));
+            continue;
         }
         if (result === "ineligible") {
-            return check("review-targets", "FAIL", `${target} has no read-only mode and cannot review.`, undefined, `Remove "${target}" from reviewRoles[].targets.`);
+            // Not a broken installation: review simply skips this target. A FAIL here
+            // would light up every config that merely still names one.
+            degraded.push(check("review-targets", "DEGRADED", `${target} has no read-only mode and is skipped when reviewing.`, undefined, `Remove "${target}" from reviewRoles[].targets.`));
+            continue;
         }
         if (result === "timeout") {
-            return check("review-targets", "FAIL", `${target} did not answer in time.`, undefined, "Re-run: agent-ops doctor --check-auth");
+            failures.push(check("review-targets", "FAIL", `${target} did not answer in time.`, undefined, "Re-run: agent-ops doctor --check-auth"));
+            continue;
         }
         if (checkAuth && result !== "ok") {
-            return check("review-targets", "FAIL", `${target} is installed but not authenticated, or it rejected the call.`, undefined, `Run: ${target} login`);
+            failures.push(check("review-targets", "FAIL", `${target} is installed but not authenticated, or it rejected the call.`, undefined, `Run: ${target} login`));
         }
     }
+    const worst = failures[0] ?? degraded[0];
+    if (worst !== undefined) {
+        return worst;
+    }
     return check("review-targets", "PASS", checkAuth
         ? `External review targets authenticated: ${targets.join(", ")}.`
         : `External review targets: ${targets.join(", ")}. ` +

package/dist/runtime/src/install/harness.js

@@ -244,7 +244,10 @@ export function managedRules(descriptor, context) {
         ""
     ];
     if (context.capabilities.includes("rules")) {
-        lines.push("For every change:", "", "1. Define two to five mechanically verifiable acceptance criteria.", "2. Inspect the smallest relevant scope and preserve unrelated changes.", "3. Apply the smallest safe change.", "4. Run evidence-producing verification for every criterion.", "5. Obtain independent review before claiming completion, via", "   `agent-ops review` (or the CLI's equivalent invocation). Never call a", "   review-target CLI (agy, codex, claude) directly — direct calls skip", "   the enforced read-only sandbox flags and can hang or fail on command", "   permission prompts.", "", "Treat `.agent-ops/config.json` as verifier authority. Discovery output is", "only a proposal until a user confirms it. Repository commands require an", "exact matching trust record. Confirmed project init/update grants it", "automatically when verification commands are configured.", "");
+        lines.push("For every change:", "", "1. Define two to five mechanically verifiable acceptance criteria.", "2. Inspect the smallest relevant scope and preserve unrelated changes.", "3. Apply the smallest safe change.", "4. Run evidence-producing verification for every criterion.", "5. Obtain independent review before claiming completion, via", "   `agent-ops review --yes` (or the CLI's equivalent invocation). Never call a", "   review-target CLI (agy, codex, claude) directly — direct calls skip", "   the enforced read-only sandbox flags and can hang or fail on command", "   permission prompts.", "", "Treat `.agent-ops/config.json` as verifier authority. Discovery output is", "only a proposal until a user confirms it. Repository commands require an", "exact matching trust record. Confirmed project init/update grants it", "automatically when verification commands are configured.", "");
+    }
+    if (context.capabilities.includes("task")) {
+        lines.push("Split a change that exceeds five acceptance criteria into subtasks:", "`agent-ops task create --parent <task-id>` records one, and", "`agent-ops task status --parent <task-id>` lists them. Each subtask", "carries its own criteria, verification, and independent review;", "completing one never completes its parent.", "");
     }
     if (context.capabilities.includes("lifecycle-summary")) {
         lines.push("Advisory lifecycle summaries and local logs are informational. Advisory", "failures must remain fail-open and cannot become verification evidence.", "");