@delete101020/aidlc 4.0.0 → 4.0.2

package/README.md

@@ -3,36 +3,32 @@
 [![License: MIT](https://img.shields.io/badge/license-MIT-97ca00)](https://github.com/delete101020/ai-native-sdlc/blob/main/LICENSE)
 [![npm](https://img.shields.io/npm/v/@delete101020/aidlc)](https://www.npmjs.com/package/@delete101020/aidlc)
 
-Terminal CLI for AIDLC — drives Claude through pipelines you declare in
-`.aidlc/workspace.yaml`. Manages the workspace, executes runs end-to-end via
-the `claude` CLI, and shares state with the VS Code extension over the
-filesystem (no daemon, no IPC).
+The terminal side of **AIDLC Native**: drive Claude through the six stages of the
+[AI-Native SDLC Playbook](https://claude.com/blog/the-ai-native-sdlc-playbook) —
+**Plan → Design → Build → Test → Deploy → Maintain** — from any shell, a CI job or
+a cron script. Every stage is an agent that writes one artifact (`intent.md`,
+`spec.md`, `plan.md`, `verify.md`, `review.md`, `incident.md`), and the
+pipeline waits at the review gates you keep.
 
-**Claude only.** The CLI shells out to `claude --print --append-system-prompt
-<skill>`. No Anthropic SDK calls, no other model runners.
+The CLI shares `.aidlc/workspace.yaml` and the run files with the
+[VS Code extension](https://marketplace.visualstudio.com/items?itemName=delete101020.aidlc-native)
+(no daemon, no IPC), so an epic started in one can be finished in the other.
 
-### New in 0.8
+**Claude only.** The CLI shells out to `claude --print --append-system-prompt
+<skill>`. No Anthropic SDK calls.
 
-`aidlc globals install` now also installs the **artifact annotation** tooling under
-`~/.claude` — a zero-dependency Markdown→HTML renderer, a vendored
-[annotron](https://www.npmjs.com/package/annotron) review editor (no global install), a
-per-epic **memory** tool, and the `/annotate-artifact` + `/epic-context` skills. Review epic
-artifacts in a browser and apply feedback back to the Markdown — from a plain terminal +
-Claude Code, no VS Code required. Optional `aidlc globals memory-hook enable` auto-loads an
-epic's memory whenever a prompt mentions it. See the [`globals`](#globals--built-in-workflow-agents--skills-under-claude) command.
+> **A fork** of [`aidlc-io/aidlc`](https://github.com/aidlc-io/aidlc) by hueanmy, not
+> affiliated with the upstream author. The package is scoped because the upstream
+> owns `aidlc` on npm; the command it installs is still `aidlc`.
 
 ## Install
 
-The CLI of **AIDLC Native**, a fork of [`aidlc-io/aidlc`](https://github.com/aidlc-io/aidlc)
-(not affiliated with the upstream author). The package is scoped because the
-upstream owns `aidlc` on npm; the command it installs is still `aidlc`:
-
 ```sh
 npm install -g @delete101020/aidlc
 aidlc --version
 ```
 
-If the upstream `aidlc` package is installed globally, `npm uninstall -g aidlc`
+If the upstream `aidlc` package is installed globally, run `npm uninstall -g aidlc`
 first — both provide the `aidlc` command.
 
 To run unreleased changes, build it from the repo:
@@ -41,7 +37,6 @@ To run unreleased changes, build it from the repo:
 pnpm install                                 # at repo root
 pnpm -r compile
 cd packages/cli && pnpm bundle && npm link   # makes `aidlc` available globally
-aidlc --version                              # 3.5.1
 ```
 
 `npm link` points at the working tree, so a later `pnpm bundle` is enough to
@@ -60,20 +55,36 @@ Run `aidlc doctor` to verify all of the above.
 ## Five-minute walkthrough
 
 ```sh
-# 1. New workspace from scratch
-mkdir my-pipeline && cd my-pipeline
+# 1. A workspace with the AI-Native SDLC workflow
+cd my-project
 aidlc init
+aidlc preset apply ai-native                 # 6 agents, 7 skills, the ai-native-full pipeline
 aidlc doctor                                 # confirm claude is wired up
 
-# 2. Drop in a built-in preset (or build manually with skill/agent/pipeline add)
-aidlc preset apply code-review
-aidlc list                                   # see the agents, skills, pipeline you got
+# 2. Start an epic — the brief picks a recipe (here: native-fix)
+aidlc epic start EPIC-1 --brief "fix the login redirect loop"
 
-# 3. Kick off a run and let Claude do the work
-aidlc run start review-pipeline --context diff=$(git diff HEAD~1)
-aidlc run exec <runId>                       # streams claude output, advances on success
+# 3. Let Claude work it, stopping at each review gate
+aidlc run exec EPIC-1                        # exit 2 = waiting for your review
+aidlc run approve EPIC-1                     # or: aidlc run reject EPIC-1 --reason "…"
+aidlc run exec EPIC-1                        # carries on to the next gate
 ```
 
+Recipes right-size the pipeline to the task:
+
+| Recipe | Steps | For |
+|---|---|---|
+| `native-full` | intent → spec → build-plan → implement → verify → review | New behaviour |
+| `native-fix` | intent → build-plan → implement → verify → review | Bugs, refactors, tech debt |
+| `native-lite` | intent → build-plan → implement → review | Small changes with a precedent in the code |
+| `native-quick` | intent → build-plan → implement → verify | Small, well-understood changes |
+| `native-align` | intent → spec | Agreeing on scope before any code |
+| `native-hotfix` | build-plan → implement → review | Production is down and the cause is known |
+
+Pick one yourself with `aidlc epic start <id> --recipe native-full`, or list them
+with `aidlc pipeline recipes`. Stage 6 has its own entry point — see
+[`maintain`](#maintain--stage-6-the-ai-native-return-path).
+
 ## Command reference
 
 Global flags available on every subcommand:
@@ -125,15 +136,16 @@ run-state JSON files are parseable. Exit 1 on any failure (including skill /
 runner / runtime checks). `--json` emits every section as
 `{ ok, failures, sections }` — a parseable CI preflight.
 
-### `mcp` — give another CLI the ast-graph server
+### `mcp` — give another CLI the code-graph server
 
 ```
 aidlc mcp status
 aidlc mcp register [--runner codex] [--dry-run]
 ```
 
-The VS Code extension registers the `ast-graph` MCP server with Claude
-automatically, project-scoped. Codex keeps MCP servers in `~/.codex/config.toml`,
+The VS Code extension registers a code-graph MCP server with Claude
+automatically, project-scoped — `ast-graph` by default, or `codegraph` when
+`aidlcNative.astGraph.engine` is set to it. Codex keeps MCP servers in `~/.codex/config.toml`,
 which is per-user: registering there points every Codex session on the machine
 at this workspace's graph, so it stays an explicit command. `register` copies
 the binary and db path out of the registration Claude already has — run
@@ -255,8 +267,9 @@ aidlc preset apply <name>                 # merges into current workspace (no ov
 aidlc preset save <name>                  # snapshot current workspace to .aidlc/presets/<name>.json
 ```
 
-Built-in presets: `code-review`, `release-notes`, `sdlc` (full 9-phase SDLC
-pipeline ported from the legacy AIDLC).
+Built-in presets: `ai-native` (the six-stage AI-Native SDLC workflow),
+`sdlc` (the classic AIDLC pipeline: Plan → (Design ∥ Test Plan) →
+Implement ∥ Generate Test Cases → Execute Test), `code-review` and `release-notes` (single-agent pipelines).
 
 ### Epics
 
@@ -265,14 +278,24 @@ under whatever `state.root` your `workspace.yaml` declares (default
 `docs/epics/`).
 
 ```
-aidlc epic list [--status pending|in_progress|done|failed] [--json]
+aidlc epic list [--status pending|in_progress|done|failed] [--tag <tag>…] [--json]
 aidlc epic status <id> [--json]                 # phase-by-phase view
 aidlc epic show <id>                            # alias for status
 aidlc epic start <id> --brief "…" [--llm]       # classify the brief → recipe → assembled pipeline
 aidlc epic start <id> --recipe <recipeId>       # or pick the recipe yourself
 aidlc epic start <id> --pipeline <pipelineId>   # or use an existing pipeline as-is
+aidlc epic start <id> … --tag payment --tag "thanh toán"   # tag it at creation
+aidlc epic tag <id>                             # print this epic's tags
+aidlc epic tag <id> payment "release q3"        # add tags (also --add/--remove/--set/--clear)
 ```
 
+**Tags** are free text in and one canonical form on disk: uppercase, ASCII,
+dash-separated — `"thanh toán VNPay"` is stored as `THANH-TOAN-VNPAY`, and the
+accents and case are folded on the way *in* to a filter too, so `--tag "thanh
+toán"` finds it. Repeating `--tag` narrows: an epic must carry every tag given.
+They live under `tags` in the epic's `state.json` and are shown, and editable,
+on each card in the VS Code panel.
+
 `epic start` mirrors the extension's smart **Start Epic**: `--brief` classifies
 the requirement into a task-type recipe (heuristic by default, `--llm` for
 model-backed classification) and assembles a right-sized pipeline.
@@ -581,12 +604,12 @@ still needed by other globally-installed workflows.
 
 ## Recipes
 
-### Drive a complete SDLC pipeline end-to-end
+### Drive an AI-Native epic end-to-end, unattended
 
 ```sh
-aidlc preset apply sdlc
-aidlc run start sdlc-pipeline --id ABC-123 --context epic=ABC-123
-aidlc run exec ABC-123 --auto-approve     # claude works through every phase
+aidlc preset apply ai-native
+aidlc epic start EPIC-7 --recipe native-full
+aidlc run exec EPIC-7 --auto-approve      # clears human gates without pausing
 ```
 
 ### Manually mark a phase done that you completed outside AIDLC
@@ -622,7 +645,7 @@ aidlc epic list --json | jq '.[] | select(.status=="in_progress") | .id'
 
 ```sh
 # Terminal 1 — kick off the run, then walk away
-aidlc run start sdlc-pipeline --id ABC-123 --context epic=ABC-123
+aidlc epic start ABC-123 --brief "…"
 aidlc run exec ABC-123 --auto-approve
 
 # Terminal 2 — live table

package/dist/bundle.js

@@ -24601,6 +24601,64 @@ var require_EpicPipelineStore = __commonJS({
   }
 });
 
+// ../core/dist/loader/epicTags.js
+var require_epicTags = __commonJS({
+  "../core/dist/loader/epicTags.js"(exports2) {
+    "use strict";
+    Object.defineProperty(exports2, "__esModule", { value: true });
+    exports2.MAX_TAG_LENGTH = exports2.EPIC_TAGS_KEY = void 0;
+    exports2.normalizeTag = normalizeTag;
+    exports2.normalizeTags = normalizeTags2;
+    exports2.readEpicTags = readEpicTags3;
+    exports2.epicMatchesTags = epicMatchesTags2;
+    exports2.applyTagEdit = applyTagEdit2;
+    exports2.EPIC_TAGS_KEY = "tags";
+    exports2.MAX_TAG_LENGTH = 48;
+    function normalizeTag(raw) {
+      if (typeof raw !== "string") {
+        return "";
+      }
+      return raw.normalize("NFD").replace(/[̀-ͯ]/g, "").replace(/[đĐ]/g, "d").toUpperCase().replace(/[^A-Z0-9_]+/g, "-").replace(/-+/g, "-").replace(/^[-_]+|[-_]+$/g, "").slice(0, exports2.MAX_TAG_LENGTH).replace(/[-_]+$/g, "");
+    }
+    function normalizeTags2(raw) {
+      const parts = Array.isArray(raw) ? raw : typeof raw === "string" ? raw.split(/[,\n]/) : [];
+      const out2 = /* @__PURE__ */ new Set();
+      for (const part of parts) {
+        const tag = normalizeTag(part);
+        if (tag) {
+          out2.add(tag);
+        }
+      }
+      return [...out2].sort();
+    }
+    function readEpicTags3(parsed) {
+      if (!parsed || typeof parsed !== "object") {
+        return [];
+      }
+      return normalizeTags2(parsed[exports2.EPIC_TAGS_KEY]);
+    }
+    function epicMatchesTags2(epicTags, filter) {
+      const wanted = normalizeTags2(filter);
+      if (wanted.length === 0) {
+        return true;
+      }
+      const have = new Set(normalizeTags2(epicTags));
+      return wanted.every((t) => have.has(t));
+    }
+    function applyTagEdit2(current, edit) {
+      const base = edit.set !== void 0 ? normalizeTags2(edit.set) : normalizeTags2(current);
+      const next = new Set(base);
+      for (const t of normalizeTags2(edit.add)) {
+        next.add(t);
+      }
+      for (const t of normalizeTags2(edit.remove)) {
+        next.delete(t);
+      }
+      return [...next].sort();
+    }
+  }
+});
+
 // ../core/dist/epics/ContextCollector.js
 var require_ContextCollector = __commonJS({
   "../core/dist/epics/ContextCollector.js"(exports2) {
@@ -24997,6 +25055,7 @@ var require_EpicScaffold = __commonJS({
     var PipelineRunner_1 = require_PipelineRunner();
     var RunStateStore_1 = require_RunStateStore();
     var EpicPipelineStore_1 = require_EpicPipelineStore();
+    var epicTags_1 = require_epicTags();
     var ContextCollector_1 = require_ContextCollector();
     var PlanGenerator_1 = require_PlanGenerator();
     var EpicScaffoldError3 = class extends Error {
@@ -25071,7 +25130,7 @@ var require_EpicScaffold = __commonJS({
       fs14.writeFileSync(stateFile, JSON.stringify(next, null, 2) + "\n", "utf8");
     }
     function scaffoldEpic2(args) {
-      const { workspaceRoot, doc, epicId, title, description, target, agents, inputs, extraProjects, pipeline, seedArtifacts, enableAutopilot = false, strictMode = true } = args;
+      const { workspaceRoot, doc, epicId, title, description, target, agents, inputs, extraProjects, pipeline, seedArtifacts, enableAutopilot = false, strictMode = true, tags } = args;
       if (!epicId.trim()) {
         throw new EpicScaffoldError3("Epic id is required.");
       }
@@ -25107,6 +25166,9 @@ var require_EpicScaffold = __commonJS({
         // because the point of a per-epic knob is that it can be found and flipped
         // by hand on an epic that turns out bigger or smaller than it looked.
         strict_mode: strictMode,
+        // Written even when empty, for the same reason as `strict_mode`: a key that
+        // is there is a key someone can find and fill in by hand.
+        [epicTags_1.EPIC_TAGS_KEY]: (0, epicTags_1.normalizeTags)(tags),
         createdAt: (/* @__PURE__ */ new Date()).toISOString(),
         stepStates: agents.map((a) => ({
           agent: a,
@@ -31818,7 +31880,7 @@ AIDLC is an AI-driven SDLC + agent workflow runner. It drives Claude through a
 pipeline you declare in \`.aidlc/workspace.yaml\`, tracking every run, step, and
 token. It has two surfaces that share the same files on disk (no daemon):
 
-- **VS Code extension** \u2014 **AIDLC Native** (\`delete101020.aidlc\`, on the VS Code
+- **VS Code extension** \u2014 **AIDLC Native** (\`delete101020.aidlc-native\`, on the VS Code
   Marketplace and Open VSX; commands and settings under \`aidlcNative.*\`) \u2014 a
   visual Builder for workspace.yaml,
   a sidebar launcher, epic/run tracking, and the **AIDLC Monitor** (token usage,
@@ -31866,7 +31928,7 @@ the other within ~200ms.
 - Bootstrap: \`aidlc init\`, \`aidlc validate\`, \`aidlc doctor\`, \`aidlc list [--json]\`.
 - Config (mirrors the Builder): \`aidlc skill|agent|pipeline add|list|show|remove\`.
 - Presets: \`aidlc preset apply|save|list\` (built-ins code-review, release-notes, sdlc).
-- Epics: \`aidlc epic list|status <id>|start <id> --brief "\u2026" [--llm]\`.
+- Epics: \`aidlc epic list [--tag <tag>]|status <id>|start <id> --brief "\u2026" [--llm]|tag <id> [tags\u2026]\`.
 - Recipes: \`aidlc recipe init\`, \`aidlc pipeline recipes|classify "<brief>"|generate\`.
 - Runs: \`aidlc run start <pipeline> --context k=v\`, \`aidlc run exec <runId> [--auto-approve]\`.
 - Watch: \`aidlc watch\`, \`aidlc tail\`, \`aidlc dashboard\`, \`aidlc monitor [--start|--open]\`.
@@ -31876,7 +31938,7 @@ the other within ~200ms.
 
 ## VS Code extension setup & UI
 1. Install **AIDLC Native** from the Extensions view (Marketplace in VS Code, Open VSX
-   in Antigravity / Cursor / VSCodium), or \`code --install-extension delete101020.aidlc\`.
+   in Antigravity / Cursor / VSCodium), or \`code --install-extension delete101020.aidlc-native\`.
    CLI: \`npm install -g @delete101020/aidlc\` (the command is \`aidlc\`).
 2. Ensure \`claude\` is on PATH and authenticated (the sidebar surfaces MCP/claude state).
 3. Open a project folder. The **AIDLC** sidebar (activity bar) shows:
@@ -31949,7 +32011,8 @@ VS Code extension over the filesystem.
 
 \u2500\u2500 Step 4: Work by epic (task-type \u2192 pipeline) \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500
   aidlc epic start <id> --brief "\u2026" [--llm]
-  aidlc epic list [--status \u2026]    aidlc epic status <id>
+  aidlc epic list [--status \u2026] [--tag \u2026]    aidlc epic status <id>
+  aidlc epic tag <id> [tags\u2026]         tag an epic (stored uppercase)
 
 \u2500\u2500 Step 5: Watch what's happening \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500
   aidlc watch          live table of all runs
@@ -31973,10 +32036,10 @@ var require_dist = __commonJS({
     Object.defineProperty(exports2, "__esModule", { value: true });
     exports2.parseFollowUps = exports2.FollowUpsParseError = exports2.FollowUpManifestSchema = exports2.FollowUpItemSchema = exports2.SIGNAL_FILE = exports2.FOLLOW_UP_ARTIFACT = exports2.readEpicSignal = exports2.existingEpicIds = exports2.followUpIdFor = exports2.followUpEpicId = exports2.renderIntentMarkdown = exports2.openFollowUpEpic = exports2.openIncidentEpic = exports2.SignalParseError = exports2.isSignal = exports2.parseSignal = exports2.SignalSchema = exports2.EpicStepEditError = exports2.resolveStepRef = exports2.describeGateEffect = exports2.commitEpicStepEdit = exports2.planSetEpicStepGates = exports2.planRemoveEpicStep = exports2.planAddEpicStep = exports2.EpicScaffoldError = exports2.epicsRoot = exports2.mapStepStatusToEpic = exports2.mirrorRunStateToEpic = exports2.scaffoldEpic = exports2.resolveArtifactCommitConfig = exports2.commitApprovedArtifacts = exports2.PipelineAdaptError = exports2.applyAdaptation = exports2.parseAdaptationVerdict = exports2.buildAdaptationPrompt = exports2.buildPhaseCatalog = exports2.slugEpicId = exports2.parseClassificationVerdict = exports2.buildClassificationPrompt = exports2.heuristicClassify = exports2.PipelineAssembleError = exports2.recipePipelineId = exports2.assemblePipeline = exports2.collectWorkspaceRefIssues = exports2.stepDagId = exports2.stepAgentId = exports2.normalizeStep = exports2.WorkspaceValidationError = exports2.validateWorkspace = exports2.WorkspaceSchema = void 0;
     exports2.composeAgentPrompt = exports2.PROJECT_INSTRUCTION_FILES = exports2.findProjectInstructions = exports2.stripPersonaMetadata = exports2.PersonaLoader = exports2.SkillNotFoundError = exports2.SkillLoader = exports2.EnvVarMissingError = exports2.EnvResolver = exports2.epicPipelineReport = exports2.planEpicPipelineExtraction = exports2.epicOwningPipeline = exports2.unstageEpicPipeline = exports2.stageEpicPipeline = exports2.writeEpicPipelines = exports2.splitEpicPipelines = exports2.mergeEpicPipelines = exports2.epicPipelinePath = exports2.EPIC_PIPELINE_FILENAME = exports2.WORKSPACE_DIR = exports2.WORKSPACE_FILENAME = exports2.WorkspaceParseError = exports2.WorkspaceNotFoundError = exports2.WorkspaceLoader = exports2.renderPlanMarkdown = exports2.generatePlan = exports2.collectContext = exports2.followUpHookFailures = exports2.pendingFollowUpDone = exports2.applyFollowUpHookRun = exports2.trackFollowUpChildren = exports2.writeFollowUpHookLedger = exports2.readFollowUpHookLedger = exports2.followUpHookLedgerPath = exports2.runFollowUpHook = exports2.fillFollowUpHookCommand = exports2.resolveFollowUpHooks = exports2.epicPipelineId = exports2.FOLLOW_UP_HOOK_TIMEOUT_MS = exports2.FOLLOW_UP_HOOKS_LEDGER = exports2.stepProducesFollowUps = exports2.collectFollowUpHookIssues = exports2.FOLLOW_UP_HOOK_PLACEHOLDERS = exports2.FOLLOW_UP_HOOK_KEYS = exports2.FOLLOW_UP_INTENT = exports2.FOLLOW_UPS_FILE = exports2.openManifestFollowUp = exports2.followUpChildId = exports2.followUpsOf = exports2.readEpicFollowUps = void 0;
-    exports2.FileRunStateStore = exports2.RunStateStore = exports2.harnessCapabilities = exports2.NO_HARNESS_CAPABILITIES = exports2.RunnerValidationError = exports2.validateRunnerExport = exports2.CustomRunnerLoader = exports2.buildClaudeSpawnEnv = exports2.hasClaudeLogin = exports2.isInsideClaudeCodeSession = exports2.createJsonSink = exports2.createLineSink = exports2.isCodexMcpConfigured = exports2.codexConfigPath = exports2.readProjectMcpServer = exports2.mcpRegistrarFor = exports2.codexMcpRegistrar = exports2.claudeMcpRegistrar = exports2.CodexRunner = exports2.DefaultRunner = exports2.RunnerRegistry = exports2.targetPath = exports2.scopePaths = exports2.discoverAssets = exports2.commandBodyIsStale = exports2.strictModeSection = exports2.resolveEpicStrictMode = exports2.epicStrictMode = exports2.commandBodyPredatesStrictMode = exports2.STRICT_MODE_KEY = exports2.STRICT_MODE_HEADING = exports2.writeUserEpicIdPrefix = exports2.userConfigPath = exports2.readUserConfig = exports2.readGitUserName = exports2.ensureUserConfigIgnored = exports2.USER_CONFIG_RELPATH = exports2.USER_CONFIG_IGNORE_LINE = exports2.suggestEpicId = exports2.resolveEpicIdPrefixChain = exports2.resolveEpicIdPrefix = exports2.epicIdDateStamp = exports2.deriveEpicIdPrefix = exports2.EPIC_ID_PREFIX_PATTERN = exports2.EPIC_ID_PREFIX_KEY = exports2.resolveArtifactLanguage = exports2.commandBodyPredatesArtifactLanguage = exports2.artifactLanguageSection = exports2.ARTIFACT_LANGUAGE_HEADING = exports2.stripPersonaDirectives = void 0;
-    exports2.getBuiltinArtifactTemplates = exports2.phaseArtifactFileName = exports2.sdlcClaudeCommand = exports2.builtinClaudeCommand = exports2.loadAllBuiltinPresets = exports2.loadBuiltinPreset = exports2.planRecipeMigration = exports2.getBuiltinRecipeSummaries = exports2.getAllBuiltinPipelineSummaries = exports2.getSdlcBuiltinPipelineSummary = exports2.getBuiltinPipelineSummary = exports2.getBuiltinWorkflowByPipelineId = exports2.getBuiltinWorkflow = exports2.workflowSlug = exports2.builtinTemplatesRoot = exports2.pipelineCommandId = exports2.PHASES = exports2.BUILTIN_WORKFLOWS = exports2.withBackfilledStepNames = exports2.describeDrift = exports2.reconcileRunSteps = exports2.RUN_STATE_SCHEMA_VERSION = exports2.migrateRunState = exports2.stepIdentity = exports2.resolvePath = exports2.AutoReviewerError = exports2.runAutoReview = exports2.renderRunReport = exports2.verifyRun = exports2.runExecLoop = exports2.estimateCostUsd = exports2.providerAliases = exports2.ratesFromConfig = exports2.mergeRates = exports2.BUILTIN_RATES = exports2.checkBudget = exports2.PipelineRunError = exports2.submitAutoReviewVerdict = exports2.requestStepUpdate = exports2.rerunStep = exports2.rejectStep = exports2.approveStep = exports2.markStepDone = exports2.canStartStep = exports2.startRun = exports2.activateBackendFromWorkspace = exports2.activateRunStateBackend = exports2.resolveRunStateBackend = exports2.GitRunStateStore = exports2.RUN_ID_PATTERN = void 0;
-    exports2.loadActiveProfile = exports2.loadProfile = exports2.listProfileManifests = exports2.resolveStandard = exports2.workspaceStandard = exports2.isBuiltinProfileId = exports2.ProfileSchema = exports2.TRACE_RULES = exports2.PROFILES_DIR = exports2.DEFAULT_PROFILE_ID = exports2.BUILTIN_PROFILE_IDS = exports2.setEpicMemoryHook = exports2.isEpicMemoryHookEnabled = exports2.installAnnotationTools = exports2.relativeEpicRoot = exports2.provisionDeclaredWorkflows = exports2.provisionWorkflowFiles = exports2.writeWorkflowArtifactTemplates = exports2.writeWorkflowCommands = exports2.provisionShortcutDocs = exports2.writeTwoLayerCommands = exports2.shortcutCommandDoc = exports2.backboneCommandDoc = exports2.unprovisionedPhases = exports2.nextEligiblePhase = exports2.resolveComposition = exports2.shortcutCommandId = exports2.getCanonicalPhase = exports2.isCanonicalPhase = exports2.BACKBONE_COMMAND_ID = exports2.CANONICAL_PHASE_IDS = exports2.CANONICAL_PHASES = exports2.renderTemplate = exports2.DEFAULT_GLOBAL_WORKFLOW_IDS = exports2.detectGlobalBuiltinSource = exports2.uninstallWorkflowGlobalsByIds = exports2.isWorkflowGloballyInstalled = exports2.installWorkflowGlobalsByIds = exports2.installGlobalDefaults = exports2.resolveProviderModel = exports2.isClaudeTierAlias = exports2.CLAUDE_TIER_ALIASES = exports2.FAST_MODEL = exports2.CODING_MODEL = exports2.PLANNING_MODEL = exports2.isBuiltinPreset = exports2.BUILTIN_PRESET_IDS = exports2.writeBuiltinAutoReviewValidators = exports2.resolvePrimaryStack = exports2.getSdlcArtifactTemplates = void 0;
-    exports2.AIDLC_CORE_VERSION = exports2.AIDLC_CLI_GUIDE_TEXT = exports2.AIDLC_KNOWLEDGE = exports2.clearResolveCommandCache = exports2.resolveCommand = exports2.remapClaudePath = exports2.claudeConfigEnv = exports2.claudeJsonPath = exports2.isDefaultClaudeConfigDir = exports2.claudeConfigDir = exports2.resolveClaudeConfigDir = exports2.defaultClaudeConfigDir = exports2.getClaudeConfigDirOverride = exports2.setClaudeConfigDir = exports2.resolveDeclaredPath = exports2.expandHome = exports2.ProfileLoadError = exports2.UnknownStandardError = exports2.builtinProfiles = void 0;
+    exports2.buildClaudeSpawnEnv = exports2.hasClaudeLogin = exports2.isInsideClaudeCodeSession = exports2.createJsonSink = exports2.createLineSink = exports2.isCodexMcpConfigured = exports2.codexConfigPath = exports2.readProjectMcpServer = exports2.mcpRegistrarFor = exports2.codexMcpRegistrar = exports2.claudeMcpRegistrar = exports2.CodexRunner = exports2.DefaultRunner = exports2.RunnerRegistry = exports2.targetPath = exports2.scopePaths = exports2.discoverAssets = exports2.commandBodyIsStale = exports2.readEpicTags = exports2.normalizeTags = exports2.normalizeTag = exports2.epicMatchesTags = exports2.applyTagEdit = exports2.MAX_TAG_LENGTH = exports2.EPIC_TAGS_KEY = exports2.strictModeSection = exports2.resolveEpicStrictMode = exports2.epicStrictMode = exports2.commandBodyPredatesStrictMode = exports2.STRICT_MODE_KEY = exports2.STRICT_MODE_HEADING = exports2.writeUserEpicIdPrefix = exports2.userConfigPath = exports2.readUserConfig = exports2.readGitUserName = exports2.ensureUserConfigIgnored = exports2.USER_CONFIG_RELPATH = exports2.USER_CONFIG_IGNORE_LINE = exports2.suggestEpicId = exports2.resolveEpicIdPrefixChain = exports2.resolveEpicIdPrefix = exports2.epicIdDateStamp = exports2.deriveEpicIdPrefix = exports2.EPIC_ID_PREFIX_PATTERN = exports2.EPIC_ID_PREFIX_KEY = exports2.resolveArtifactLanguage = exports2.commandBodyPredatesArtifactLanguage = exports2.artifactLanguageSection = exports2.ARTIFACT_LANGUAGE_HEADING = exports2.stripPersonaDirectives = void 0;
+    exports2.getBuiltinRecipeSummaries = exports2.getAllBuiltinPipelineSummaries = exports2.getSdlcBuiltinPipelineSummary = exports2.getBuiltinPipelineSummary = exports2.getBuiltinWorkflowByPipelineId = exports2.getBuiltinWorkflow = exports2.workflowSlug = exports2.builtinTemplatesRoot = exports2.pipelineCommandId = exports2.PHASES = exports2.BUILTIN_WORKFLOWS = exports2.withBackfilledStepNames = exports2.describeDrift = exports2.reconcileRunSteps = exports2.RUN_STATE_SCHEMA_VERSION = exports2.migrateRunState = exports2.stepIdentity = exports2.resolvePath = exports2.AutoReviewerError = exports2.runAutoReview = exports2.renderRunReport = exports2.verifyRun = exports2.runExecLoop = exports2.estimateCostUsd = exports2.providerAliases = exports2.ratesFromConfig = exports2.mergeRates = exports2.BUILTIN_RATES = exports2.checkBudget = exports2.PipelineRunError = exports2.submitAutoReviewVerdict = exports2.requestStepUpdate = exports2.rerunStep = exports2.rejectStep = exports2.approveStep = exports2.markStepDone = exports2.canStartStep = exports2.startRun = exports2.activateBackendFromWorkspace = exports2.activateRunStateBackend = exports2.resolveRunStateBackend = exports2.GitRunStateStore = exports2.RUN_ID_PATTERN = exports2.FileRunStateStore = exports2.RunStateStore = exports2.harnessCapabilities = exports2.NO_HARNESS_CAPABILITIES = exports2.RunnerValidationError = exports2.validateRunnerExport = exports2.CustomRunnerLoader = void 0;
+    exports2.TRACE_RULES = exports2.PROFILES_DIR = exports2.DEFAULT_PROFILE_ID = exports2.BUILTIN_PROFILE_IDS = exports2.setEpicMemoryHook = exports2.isEpicMemoryHookEnabled = exports2.installAnnotationTools = exports2.relativeEpicRoot = exports2.provisionDeclaredWorkflows = exports2.provisionWorkflowFiles = exports2.writeWorkflowArtifactTemplates = exports2.writeWorkflowCommands = exports2.provisionShortcutDocs = exports2.writeTwoLayerCommands = exports2.shortcutCommandDoc = exports2.backboneCommandDoc = exports2.unprovisionedPhases = exports2.nextEligiblePhase = exports2.resolveComposition = exports2.shortcutCommandId = exports2.getCanonicalPhase = exports2.isCanonicalPhase = exports2.BACKBONE_COMMAND_ID = exports2.CANONICAL_PHASE_IDS = exports2.CANONICAL_PHASES = exports2.renderTemplate = exports2.DEFAULT_GLOBAL_WORKFLOW_IDS = exports2.detectGlobalBuiltinSource = exports2.uninstallWorkflowGlobalsByIds = exports2.isWorkflowGloballyInstalled = exports2.installWorkflowGlobalsByIds = exports2.installGlobalDefaults = exports2.resolveProviderModel = exports2.isClaudeTierAlias = exports2.CLAUDE_TIER_ALIASES = exports2.FAST_MODEL = exports2.CODING_MODEL = exports2.PLANNING_MODEL = exports2.isBuiltinPreset = exports2.BUILTIN_PRESET_IDS = exports2.writeBuiltinAutoReviewValidators = exports2.resolvePrimaryStack = exports2.getSdlcArtifactTemplates = exports2.getBuiltinArtifactTemplates = exports2.phaseArtifactFileName = exports2.sdlcClaudeCommand = exports2.builtinClaudeCommand = exports2.loadAllBuiltinPresets = exports2.loadBuiltinPreset = exports2.planRecipeMigration = void 0;
+    exports2.AIDLC_CORE_VERSION = exports2.AIDLC_CLI_GUIDE_TEXT = exports2.AIDLC_KNOWLEDGE = exports2.clearResolveCommandCache = exports2.resolveCommand = exports2.remapClaudePath = exports2.claudeConfigEnv = exports2.claudeJsonPath = exports2.isDefaultClaudeConfigDir = exports2.claudeConfigDir = exports2.resolveClaudeConfigDir = exports2.defaultClaudeConfigDir = exports2.getClaudeConfigDirOverride = exports2.setClaudeConfigDir = exports2.resolveDeclaredPath = exports2.expandHome = exports2.ProfileLoadError = exports2.UnknownStandardError = exports2.builtinProfiles = exports2.loadActiveProfile = exports2.loadProfile = exports2.listProfileManifests = exports2.resolveStandard = exports2.workspaceStandard = exports2.isBuiltinProfileId = exports2.ProfileSchema = void 0;
     var WorkspaceSchema_1 = require_WorkspaceSchema();
     Object.defineProperty(exports2, "WorkspaceSchema", { enumerable: true, get: function() {
       return WorkspaceSchema_1.WorkspaceSchema;
@@ -32377,6 +32440,28 @@ var require_dist = __commonJS({
     Object.defineProperty(exports2, "strictModeSection", { enumerable: true, get: function() {
       return strictMode_1.strictModeSection;
     } });
+    var epicTags_1 = require_epicTags();
+    Object.defineProperty(exports2, "EPIC_TAGS_KEY", { enumerable: true, get: function() {
+      return epicTags_1.EPIC_TAGS_KEY;
+    } });
+    Object.defineProperty(exports2, "MAX_TAG_LENGTH", { enumerable: true, get: function() {
+      return epicTags_1.MAX_TAG_LENGTH;
+    } });
+    Object.defineProperty(exports2, "applyTagEdit", { enumerable: true, get: function() {
+      return epicTags_1.applyTagEdit;
+    } });
+    Object.defineProperty(exports2, "epicMatchesTags", { enumerable: true, get: function() {
+      return epicTags_1.epicMatchesTags;
+    } });
+    Object.defineProperty(exports2, "normalizeTag", { enumerable: true, get: function() {
+      return epicTags_1.normalizeTag;
+    } });
+    Object.defineProperty(exports2, "normalizeTags", { enumerable: true, get: function() {
+      return epicTags_1.normalizeTags;
+    } });
+    Object.defineProperty(exports2, "readEpicTags", { enumerable: true, get: function() {
+      return epicTags_1.readEpicTags;
+    } });
     var commandBodyFreshness_1 = require_commandBodyFreshness();
     Object.defineProperty(exports2, "commandBodyIsStale", { enumerable: true, get: function() {
       return commandBodyFreshness_1.commandBodyIsStale;
@@ -41510,7 +41595,7 @@ var require_package = __commonJS({
   "package.json"(exports2, module2) {
     module2.exports = {
       name: "@delete101020/aidlc",
-      version: "4.0.0",
+      version: "4.0.2",
       description: "AIDLC Native CLI, a fork of AIDLC \u2014 drive Claude through the six-stage AI-Native SDLC Playbook: Plan \u2192 Design \u2192 Build \u2192 Test \u2192 Deploy \u2192 Maintain, with artifact gates, requirements traceability, an incident loop back to Plan, and Spec Kit support.",
       license: "MIT",
       repository: {
@@ -44179,7 +44264,7 @@ var safeLoadAll = renamed("safeLoadAll", "loadAll");
 var safeDump = renamed("safeDump", "dump");
 
 // src/index.ts
-var import_core25 = __toESM(require_dist());
+var import_core26 = __toESM(require_dist());
 
 // src/commands/validate.ts
 var import_chalk = __toESM(require_source());
@@ -47130,11 +47215,12 @@ var http = __toESM(require("http"));
 var path11 = __toESM(require("path"));
 var import_chalk14 = __toESM(require_source());
 var import_chokidar3 = __toESM(require_chokidar());
-var import_core17 = __toESM(require_dist());
+var import_core18 = __toESM(require_dist());
 
 // src/epicsList.ts
 var fs9 = __toESM(require("fs"));
 var path10 = __toESM(require("path"));
+var import_core17 = __toESM(require_dist());
 var STATUS_VALUES = ["pending", "in_progress", "done", "failed"];
 function asStatus(v) {
   return STATUS_VALUES.includes(v) ? v : "pending";
@@ -47178,6 +47264,7 @@ function listEpics(workspaceRoot, doc) {
       description: typeof parsed.description === "string" ? parsed.description : "",
       status: asStatus(parsed.status),
       createdAt: typeof parsed.createdAt === "string" ? parsed.createdAt : "",
+      tags: (0, import_core17.readEpicTags)(parsed),
       pipeline: typeof parsed.pipeline === "string" ? parsed.pipeline : null,
       agents: Array.isArray(parsed.agents) ? parsed.agents.map(String) : [],
       currentStep: typeof parsed.currentStep === "number" ? parsed.currentStep : 0,
@@ -47234,12 +47321,12 @@ function registerDashboard(program3) {
       }
       if (url.pathname === "/api/runs") {
         res.writeHead(200, { "Content-Type": "application/json" });
-        res.end(JSON.stringify(import_core17.RunStateStore.list(root)));
+        res.end(JSON.stringify(import_core18.RunStateStore.list(root)));
         return;
       }
       if (url.pathname.startsWith("/api/runs/")) {
         const runId = url.pathname.slice("/api/runs/".length);
-        const state = import_core17.RunStateStore.load(root, runId);
+        const state = import_core18.RunStateStore.load(root, runId);
         if (!state) {
           res.writeHead(404);
           res.end("not found");
@@ -47381,7 +47468,7 @@ function handleAction(req, res, root) {
       res.end(JSON.stringify({ error: "invalid JSON" }));
       return;
     }
-    const state = import_core17.RunStateStore.load(root, payload.runId);
+    const state = import_core18.RunStateStore.load(root, payload.runId);
     if (!state) {
       res.writeHead(404);
       res.end(JSON.stringify({ error: `run not found: ${payload.runId}` }));
@@ -47391,31 +47478,31 @@ function handleAction(req, res, root) {
     try {
       switch (payload.type) {
         case "mark-done": {
-          const ws = import_core17.WorkspaceLoader.load(root);
+          const ws = import_core18.WorkspaceLoader.load(root);
           const pipeline = ws.config.pipelines.find((p) => p.id === state.pipelineId);
           if (!pipeline) {
             throw new Error(`pipeline ${state.pipelineId} not found`);
           }
-          next = (0, import_core17.markStepDone)({ state, pipeline, workspaceRoot: root });
+          next = (0, import_core18.markStepDone)({ state, pipeline, workspaceRoot: root });
           break;
         }
         case "approve": {
-          const ws = import_core17.WorkspaceLoader.load(root);
+          const ws = import_core18.WorkspaceLoader.load(root);
           const pipeline = ws.config.pipelines.find((p) => p.id === state.pipelineId);
           if (!pipeline) {
             throw new Error(`pipeline ${state.pipelineId} not found`);
           }
-          next = (0, import_core17.approveStep)({ state, pipeline });
+          next = (0, import_core18.approveStep)({ state, pipeline });
           if (payload.comment) {
             next.steps[state.currentStepIdx].feedback = payload.comment;
           }
           break;
         }
         case "reject":
-          next = (0, import_core17.rejectStep)({ state, reason: payload.reason ?? "" });
+          next = (0, import_core18.rejectStep)({ state, reason: payload.reason ?? "" });
           break;
         case "rerun":
-          next = (0, import_core17.rerunStep)({ state, feedback: payload.feedback });
+          next = (0, import_core18.rerunStep)({ state, feedback: payload.feedback });
           break;
         default:
           res.writeHead(400);
@@ -47424,7 +47511,7 @@ function handleAction(req, res, root) {
       }
     } catch (err) {
       const msg2 = err instanceof Error ? err.message : String(err);
-      const missing = err instanceof import_core17.PipelineRunError ? err.missing : void 0;
+      const missing = err instanceof import_core18.PipelineRunError ? err.missing : void 0;
       res.writeHead(400);
       res.end(JSON.stringify({ error: msg2, missing }));
       return;
@@ -47963,27 +48050,39 @@ var import_chalk15 = __toESM(require_source());
 var import_cli_table32 = __toESM(require_cli_table3());
 var fs10 = __toESM(require("fs"));
 var path12 = __toESM(require("path"));
-var import_core18 = __toESM(require_dist());
+var import_core19 = __toESM(require_dist());
 function registerEpic(program3) {
   const cmd = program3.command("epic").description("List + inspect epics from <state.root>/<id>/state.json (mirrors the extension)");
-  cmd.command("list").description("List all epics found under workspace state.root (default: docs/epics/)").option("--json", "Output raw JSON").option("--status <status>", "Filter by status (pending | in_progress | done | failed)").action((opts, actionCmd) => {
+  cmd.command("list").description("List all epics found under workspace state.root (default: docs/epics/)").option("--json", "Output raw JSON").option("--status <status>", "Filter by status (pending | in_progress | done | failed)").option(
+    "--tag <tag>",
+    'Filter by tag \u2014 repeatable, and an epic must carry all of them. Case and accents are folded, so --tag "thanh to\xE1n" finds THANH-TOAN',
+    collectKv,
+    []
+  ).action((opts, actionCmd) => {
     const root = resolveWorkspaceRoot(actionCmd);
     const doc = readYaml(root);
     let epics = listEpics(root, doc);
     if (opts.status) {
       epics = epics.filter((e) => e.status === opts.status);
     }
+    const wantedTags = (0, import_core19.normalizeTags)(opts.tag);
+    if (wantedTags.length > 0) {
+      epics = epics.filter((e) => (0, import_core19.epicMatchesTags)(e.tags, wantedTags));
+    }
     if (opts.json) {
       console.log(JSON.stringify(epics, null, 2));
       return;
     }
     if (epics.length === 0) {
       console.log(import_chalk15.default.dim("No epics found."));
+      if (wantedTags.length > 0) {
+        console.log(import_chalk15.default.dim(`  tags = ${wantedTags.join(" + ")}`));
+      }
       console.log(import_chalk15.default.dim(`  state.root = ${doc?.state ? doc.state.root ?? "docs/epics" : "docs/epics"}`));
       return;
     }
     const table = new import_cli_table32.default({
-      head: [import_chalk15.default.bold("Epic"), import_chalk15.default.bold("Title"), import_chalk15.default.bold("Progress"), import_chalk15.default.bold("Status"), import_chalk15.default.bold("Pipeline")],
+      head: [import_chalk15.default.bold("Epic"), import_chalk15.default.bold("Title"), import_chalk15.default.bold("Tags"), import_chalk15.default.bold("Progress"), import_chalk15.default.bold("Status"), import_chalk15.default.bold("Pipeline")],
       style: { head: [], border: [] }
     });
     for (const epic of epics) {
@@ -47994,6 +48093,7 @@ function registerEpic(program3) {
       table.push([
         import_chalk15.default.bold(epic.id),
         truncate(epic.title || import_chalk15.default.dim("(untitled)"), 40),
+        epic.tags.length > 0 ? import_chalk15.default.cyan(truncate(epic.tags.join(" "), 28)) : import_chalk15.default.dim("\u2014"),
         stepLabel2,
         colorEpicStatus(epic.status),
         import_chalk15.default.dim(epic.pipeline ?? "\u2014")
@@ -48021,7 +48121,7 @@ ${epics.length} epic${epics.length !== 1 ? "s" : ""}`));
     }
     printEpicDetail(epic);
   });
-  cmd.command("start <epicId>").description(`Scaffold a new epic on disk (folder + artifacts + run state) \u2014 mirrors the extension's "Start epic"`).option("--recipe <id>", "assemble a right-sized pipeline from this recipe").option("--pipeline <id>", "use an existing pipeline as-is").option("--brief <text...>", "classify this requirement brief into a recipe, then assemble").option("--llm", "use the `claude` CLI to classify --brief (falls back to heuristic)").option("--from <pipelineId>", "override the recipe's source pipeline").option("--title <title>", "epic title").option("--desc <description>", "epic description / requirement snapshot").option("--input <kv>", "capability input as key=value (repeatable)", collectKv, []).option("--no-strict", "work to the size of this epic \u2014 no invented NFR / risk / alternatives sections").action((epicId, opts, actionCmd) => {
+  cmd.command("start <epicId>").description(`Scaffold a new epic on disk (folder + artifacts + run state) \u2014 mirrors the extension's "Start epic"`).option("--recipe <id>", "assemble a right-sized pipeline from this recipe").option("--pipeline <id>", "use an existing pipeline as-is").option("--brief <text...>", "classify this requirement brief into a recipe, then assemble").option("--llm", "use the `claude` CLI to classify --brief (falls back to heuristic)").option("--from <pipelineId>", "override the recipe's source pipeline").option("--title <title>", "epic title").option("--desc <description>", "epic description / requirement snapshot").option("--input <kv>", "capability input as key=value (repeatable)", collectKv, []).option("--tag <tag>", 'tag for this epic (repeatable) \u2014 stored uppercase, e.g. "thanh to\xE1n" \u2192 THANH-TOAN', collectKv, []).option("--no-strict", "work to the size of this epic \u2014 no invented NFR / risk / alternatives sections").action((epicId, opts, actionCmd) => {
     const root = resolveWorkspaceRoot(actionCmd);
     const doc = requireYaml(root);
     const modes = [opts.recipe, opts.pipeline, opts.brief?.length ? "brief" : void 0].filter(Boolean).length;
@@ -48031,7 +48131,7 @@ ${epics.length} epic${epics.length !== 1 ? "s" : ""}`));
     }
     let config;
     try {
-      config = (0, import_core18.validateWorkspace)(doc, ".aidlc/workspace.yaml");
+      config = (0, import_core19.validateWorkspace)(doc, ".aidlc/workspace.yaml");
     } catch (err) {
       console.error(import_chalk15.default.red("workspace.yaml is invalid \u2014 fix it before starting an epic:"));
       console.error(import_chalk15.default.dim(err instanceof Error ? err.message : String(err)));
@@ -48055,7 +48155,7 @@ ${epics.length} epic${epics.length !== 1 ? "s" : ""}`));
       let recipeId = opts.recipe;
       if (!recipeId) {
         const brief = (opts.brief ?? []).join(" ").trim();
-        const heur = (0, import_core18.heuristicClassify)(brief, config.recipes);
+        const heur = (0, import_core19.heuristicClassify)(brief, config.recipes);
         let verdict = heur;
         if (opts.llm) {
           console.log(import_chalk15.default.dim(`Provisional \u2192 ${heur.recipeId} (${heur.confidence}, heuristic) \u2014 refining with claude\u2026`));
@@ -48070,20 +48170,20 @@ ${epics.length} epic${epics.length !== 1 ? "s" : ""}`));
           recipe.from = opts.from;
         }
       }
-      const pipelineId = (0, import_core18.recipePipelineId)({ recipeId, epicId, taken: existingIds(doc.pipelines) });
+      const pipelineId = (0, import_core19.recipePipelineId)({ recipeId, epicId, taken: existingIds(doc.pipelines) });
       try {
-        pipelineCfg = (0, import_core18.assemblePipeline)(config, { recipeId, pipelineId });
+        pipelineCfg = (0, import_core19.assemblePipeline)(config, { recipeId, pipelineId });
       } catch (err) {
-        if (err instanceof import_core18.PipelineAssembleError) {
+        if (err instanceof import_core19.PipelineAssembleError) {
           console.error(import_chalk15.default.red("Could not assemble pipeline: ") + import_chalk15.default.dim(err.message));
           process.exit(1);
         }
         throw err;
       }
       doc.pipelines.push(pipelineCfg);
-      (0, import_core18.stageEpicPipeline)(doc, pipelineCfg.id, epicId);
+      (0, import_core19.stageEpicPipeline)(doc, pipelineCfg.id, epicId);
       try {
-        (0, import_core18.validateWorkspace)(doc, ".aidlc/workspace.yaml");
+        (0, import_core19.validateWorkspace)(doc, ".aidlc/workspace.yaml");
       } catch (err) {
         console.error(import_chalk15.default.red("Assembled pipeline failed validation \u2014 not written:"));
         console.error(import_chalk15.default.dim(err instanceof Error ? err.message : String(err)));
@@ -48092,7 +48192,7 @@ ${epics.length} epic${epics.length !== 1 ? "s" : ""}`));
       writeYaml(root, doc);
       console.log(import_chalk15.default.dim(`Assembled pipeline ${import_chalk15.default.bold(pipelineCfg.id)} from recipe ${import_chalk15.default.bold(recipeId)}`));
     }
-    const agents = Array.isArray(pipelineCfg.steps) ? pipelineCfg.steps.map(import_core18.stepAgentId) : [];
+    const agents = Array.isArray(pipelineCfg.steps) ? pipelineCfg.steps.map(import_core19.stepAgentId) : [];
     const inputs = {};
     for (const kv of opts.input) {
       const eq = kv.indexOf("=");
@@ -48101,7 +48201,7 @@ ${epics.length} epic${epics.length !== 1 ? "s" : ""}`));
       }
     }
     try {
-      const { epicDir } = (0, import_core18.scaffoldEpic)({
+      const { epicDir } = (0, import_core19.scaffoldEpic)({
         workspaceRoot: root,
         doc,
         epicId,
@@ -48111,13 +48211,18 @@ ${epics.length} epic${epics.length !== 1 ? "s" : ""}`));
         agents,
         inputs,
         pipeline: pipelineCfg,
-        strictMode: opts.strict
+        strictMode: opts.strict,
+        tags: opts.tag
       });
+      const tags = (0, import_core19.normalizeTags)(opts.tag);
       const steps = agents.join(" \u2192 ");
       console.log(import_chalk15.default.green("\u2714") + ` Started epic ${import_chalk15.default.bold(epicId)}`);
       console.log(import_chalk15.default.dim(`  Pipeline: ${pipelineCfg.id}`));
       console.log(import_chalk15.default.dim(`  Steps:    ${steps}`));
       console.log(import_chalk15.default.dim(`  Dir:      ${epicDir}`));
+      if (tags.length > 0) {
+        console.log(import_chalk15.default.dim("  Tags:     ") + import_chalk15.default.cyan(tags.join(" ")));
+      }
       if (!opts.strict) {
         console.log(import_chalk15.default.dim("  Depth:    strict_mode: false \u2014 phases stay proportional to the work"));
       }
@@ -48131,20 +48236,79 @@ ${epics.length} epic${epics.length !== 1 ? "s" : ""}`));
       console.log(`
 Run ${import_chalk15.default.cyan(`${runCmd} ${epicId}`)} in Claude to begin.`);
     } catch (err) {
-      if (err instanceof import_core18.EpicScaffoldError) {
+      if (err instanceof import_core19.EpicScaffoldError) {
         console.error(import_chalk15.default.red(err.message));
         process.exit(1);
       }
       throw err;
     }
   });
+  cmd.command("tag <id> [tags...]").description("Show or edit an epic's tags. Bare `tag <id>` prints them; extra words are added.").option("--add <tag>", "add a tag (repeatable) \u2014 same as passing it positionally", collectKv, []).option("--remove <tag>", "remove a tag (repeatable)", collectKv, []).option("--set <tag>", "replace every tag with these (repeatable)", collectKv, []).option("--clear", "remove every tag").option("--json", "Output the resulting tags as JSON").action((id, positional, opts, actionCmd) => {
+    const root = resolveWorkspaceRoot(actionCmd);
+    const doc = readYaml(root);
+    const file = path12.join((0, import_core19.epicsRoot)(root, doc), id, "state.json");
+    if (!fs10.existsSync(file)) {
+      const all = listEpics(root, doc).map((e) => e.id);
+      console.error(import_chalk15.default.red(`Epic "${id}" not found.`));
+      if (all.length > 0) {
+        console.error(import_chalk15.default.dim(`Available: ${all.join(", ")}`));
+      }
+      process.exit(1);
+    }
+    let state;
+    try {
+      state = JSON.parse(fs10.readFileSync(file, "utf8"));
+    } catch (err) {
+      console.error(import_chalk15.default.red(`Could not read ${file}: ${err instanceof Error ? err.message : String(err)}`));
+      process.exit(1);
+      return;
+    }
+    const before = (0, import_core19.readEpicTags)(state);
+    const add = [...positional, ...opts.add];
+    const clear = opts.clear === true;
+    const hasEdit = clear || add.length > 0 || opts.remove.length > 0 || opts.set.length > 0;
+    if (!hasEdit) {
+      if (opts.json) {
+        console.log(JSON.stringify(before, null, 2));
+        return;
+      }
+      console.log(before.length > 0 ? import_chalk15.default.cyan(before.join(" ")) : import_chalk15.default.dim("(no tags)"));
+      return;
+    }
+    const after = (0, import_core19.applyTagEdit)(before, {
+      // `--clear` is `--set` with nothing in it; spelling it separately keeps
+      // "remove everything" from needing an empty-string argument.
+      set: clear ? [] : opts.set.length > 0 ? opts.set : void 0,
+      add,
+      remove: opts.remove
+    });
+    if (after.join("\0") === before.join("\0")) {
+      console.log(import_chalk15.default.dim(`No change \u2014 ${id} already has ${before.length > 0 ? import_chalk15.default.cyan(before.join(" ")) : "no tags"}.`));
+      return;
+    }
+    state[import_core19.EPIC_TAGS_KEY] = after;
+    try {
+      fs10.writeFileSync(file, JSON.stringify(state, null, 2) + "\n", "utf8");
+    } catch (err) {
+      console.error(import_chalk15.default.red(`Could not write ${file}: ${err instanceof Error ? err.message : String(err)}`));
+      process.exit(1);
+    }
+    if (opts.json) {
+      console.log(JSON.stringify(after, null, 2));
+      return;
+    }
+    console.log(import_chalk15.default.green("\u2714") + ` ${import_chalk15.default.bold(id)} tags: ` + (after.length > 0 ? import_chalk15.default.cyan(after.join(" ")) : import_chalk15.default.dim("(none)")));
+    if (before.length > 0) {
+      console.log(import_chalk15.default.dim(`  was: ${before.join(" ")}`));
+    }
+  });
   const pipelineCmd = cmd.command("pipeline").description("Move epic-owned pipelines out of the shared workspace.yaml");
   pipelineCmd.command("extract").description(
     "Move each epic's pipeline out of .aidlc/workspace.yaml into\n  <state.root>/<epic>/pipeline.yaml. Only pipelines an epic names in its\n  own state.json move; shared and hand-authored ones stay put."
   ).option("--dry-run", "List what would move, write nothing").action((opts, actionCmd) => {
     const root = resolveWorkspaceRoot(actionCmd);
     const doc = requireYaml(root);
-    const plan = (0, import_core18.planEpicPipelineExtraction)(root, doc);
+    const plan = (0, import_core19.planEpicPipelineExtraction)(root, doc);
     if (plan.length === 0) {
       console.log(import_chalk15.default.dim("No inline epic pipelines left to move."));
       return;
@@ -48160,7 +48324,7 @@ ${plan.length} pipeline(s) would move. Re-run without --dry-run.`));
       return;
     }
     for (const item of plan) {
-      (0, import_core18.stageEpicPipeline)(doc, item.pipelineId, item.epicId);
+      (0, import_core19.stageEpicPipeline)(doc, item.pipelineId, item.epicId);
     }
     writeYaml(root, doc);
     console.log(import_chalk15.default.green("\u2714") + ` Moved ${plan.length} pipeline(s) out of .aidlc/workspace.yaml.`);
@@ -48169,19 +48333,19 @@ ${plan.length} pipeline(s) would move. Re-run without --dry-run.`));
   cmd.command("next-id").description("Print the id to give the next epic \u2014 honours epic_id_prefix in .aidlc/user.yaml").action((_opts, actionCmd) => {
     const root = resolveWorkspaceRoot(actionCmd);
     const doc = requireYaml(root);
-    const dir = (0, import_core18.epicsRoot)(root, doc);
+    const dir = (0, import_core19.epicsRoot)(root, doc);
     const existing = fs10.existsSync(dir) ? fs10.readdirSync(dir, { withFileTypes: true }).filter((d) => d.isDirectory()).map((d) => d.name) : [];
-    const { prefix } = (0, import_core18.resolveEpicIdPrefixChain)({
-      user: (0, import_core18.readUserConfig)(root),
+    const { prefix } = (0, import_core19.resolveEpicIdPrefixChain)({
+      user: (0, import_core19.readUserConfig)(root),
       workspace: doc,
-      gitUserName: (0, import_core18.readGitUserName)(root)
+      gitUserName: (0, import_core19.readGitUserName)(root)
     });
-    console.log((0, import_core18.suggestEpicId)(existing, prefix));
+    console.log((0, import_core19.suggestEpicId)(existing, prefix));
   });
   cmd.command("strict <epicId> [value]").description("Show or set an epic's depth of work (strict_mode in its state.json)").action((epicId, value, _opts, actionCmd) => {
     const root = resolveWorkspaceRoot(actionCmd);
     const doc = requireYaml(root);
-    const file = path12.join((0, import_core18.epicsRoot)(root, doc), epicId, "state.json");
+    const file = path12.join((0, import_core19.epicsRoot)(root, doc), epicId, "state.json");
     if (!fs10.existsSync(file)) {
       console.error(import_chalk15.default.red(`No epic "${epicId}" \u2014 expected ${file}`));
       process.exit(1);
@@ -48195,7 +48359,7 @@ ${plan.length} pipeline(s) would move. Re-run without --dry-run.`));
       return;
     }
     if (value === void 0) {
-      const on = (0, import_core18.epicStrictMode)(state);
+      const on = (0, import_core19.epicStrictMode)(state);
       console.log(`${epicId}: strict_mode ${on ? import_chalk15.default.bold("true") : import_chalk15.default.bold("false")}`);
       console.log(import_chalk15.default.dim(on ? "  Phases work at full depth. Set false for an epic the size of one task." : "  Phases stay proportional to the work \u2014 no invented NFR / risk sections."));
       return;
@@ -48207,7 +48371,7 @@ ${plan.length} pipeline(s) would move. Re-run without --dry-run.`));
       console.error(import_chalk15.default.red(`Expected on|off, got "${value}".`));
       process.exit(1);
     }
-    state[import_core18.STRICT_MODE_KEY] = truthy.includes(v);
+    state[import_core19.STRICT_MODE_KEY] = truthy.includes(v);
     fs10.writeFileSync(file, JSON.stringify(state, null, 2) + "\n", "utf8");
     console.log(import_chalk15.default.green("\u2714") + ` ${epicId}: strict_mode ${truthy.includes(v)}`);
     console.log(import_chalk15.default.dim("  Takes effect on the next phase run \u2014 nothing already written changes."));
@@ -48222,7 +48386,7 @@ ${plan.length} pipeline(s) would move. Re-run without --dry-run.`));
     const root = resolveWorkspaceRoot(actionCmd);
     const doc = requireYaml(root);
     const { runState, pipelineCfg } = requireEditableEpic(root, doc, epicId);
-    const plan = runEdit(() => (0, import_core18.planAddEpicStep)({
+    const plan = runEdit(() => (0, import_core19.planAddEpicStep)({
       runState,
       pipeline: pipelineCfg,
       step: {
@@ -48247,7 +48411,7 @@ ${plan.length} pipeline(s) would move. Re-run without --dry-run.`));
     const root = resolveWorkspaceRoot(actionCmd);
     const doc = requireYaml(root);
     const { runState, pipelineCfg } = requireEditableEpic(root, doc, epicId);
-    const plan = runEdit(() => (0, import_core18.planRemoveEpicStep)({ runState, pipeline: pipelineCfg, step }));
+    const plan = runEdit(() => (0, import_core19.planRemoveEpicStep)({ runState, pipeline: pipelineCfg, step }));
     writeEdit(root, doc, plan);
     console.log(import_chalk15.default.green("\u2714") + ` Removed step ${import_chalk15.default.bold(plan.stepId)} from ${import_chalk15.default.bold(pipelineCfg.id)}`);
     printSteps(plan);
@@ -48256,7 +48420,7 @@ ${plan.length} pipeline(s) would move. Re-run without --dry-run.`));
     const root = resolveWorkspaceRoot(actionCmd);
     const doc = requireYaml(root);
     const { runState, pipelineCfg } = requireEditableEpic(root, doc, epicId);
-    const plan = runEdit(() => (0, import_core18.planSetEpicStepGates)({
+    const plan = runEdit(() => (0, import_core19.planSetEpicStepGates)({
       runState,
       pipeline: pipelineCfg,
       step,
@@ -48276,14 +48440,14 @@ ${plan.length} pipeline(s) would move. Re-run without --dry-run.`));
       const arrow = c.to ? import_chalk15.default.yellow("on") : import_chalk15.default.dim("off");
       console.log(`  ${import_chalk15.default.dim(c.gate)} ${import_chalk15.default.dim(String(c.from))} \u2192 ${arrow}`);
     }
-    const note = (0, import_core18.describeGateEffect)(plan.stepStatus, plan.changes);
+    const note = (0, import_core19.describeGateEffect)(plan.stepStatus, plan.changes);
     if (note) {
       console.log(import_chalk15.default.yellow(`  The step is ${plan.stepStatus}. `) + import_chalk15.default.dim(note));
     }
   });
 }
 function requireEditableEpic(root, doc, epicId) {
-  const runState = import_core18.RunStateStore.load(root, epicId);
+  const runState = import_core19.RunStateStore.load(root, epicId);
   if (!runState) {
     console.error(import_chalk15.default.red(`Epic "${epicId}" has no run state at .aidlc/runs/${epicId}.json.`));
     console.error(import_chalk15.default.dim("  Only a pipeline-backed epic has a step list to edit."));
@@ -48307,7 +48471,7 @@ function runEdit(plan) {
   try {
     return plan();
   } catch (err) {
-    if (err instanceof import_core18.EpicStepEditError) {
+    if (err instanceof import_core19.EpicStepEditError) {
       console.error(import_chalk15.default.red(err.message));
       process.exit(1);
     }
@@ -48316,7 +48480,7 @@ function runEdit(plan) {
 }
 function writeEdit(root, doc, plan) {
   const before = JSON.parse(JSON.stringify(doc.pipelines));
-  (0, import_core18.commitEpicStepEdit)({
+  (0, import_core19.commitEpicStepEdit)({
     workspaceRoot: root,
     doc,
     plan,
@@ -48335,13 +48499,13 @@ function writePipeline(root, doc, pipeline) {
   );
   let config;
   try {
-    config = (0, import_core18.validateWorkspace)(doc, ".aidlc/workspace.yaml");
+    config = (0, import_core19.validateWorkspace)(doc, ".aidlc/workspace.yaml");
   } catch (err) {
     console.error(import_chalk15.default.red("The edited pipeline fails workspace validation \u2014 nothing was written:"));
     console.error(import_chalk15.default.dim(err instanceof Error ? err.message : String(err)));
     process.exit(1);
   }
-  const issues = (0, import_core18.collectWorkspaceRefIssues)(config).filter((i) => i.path.startsWith(`pipelines.${pipeline.id}.`));
+  const issues = (0, import_core19.collectWorkspaceRefIssues)(config).filter((i) => i.path.startsWith(`pipelines.${pipeline.id}.`));
   if (issues.length > 0) {
     console.error(import_chalk15.default.red("The edited step references something the workspace does not define \u2014 nothing was written:"));
     for (const issue of issues) {
@@ -48355,7 +48519,7 @@ function printSteps(plan) {
   const { runState } = plan;
   runState.steps.forEach((s, i) => {
     const marker = i === runState.currentStepIdx ? import_chalk15.default.yellow("\u25B6") : " ";
-    const label = i === plan.index ? import_chalk15.default.bold((0, import_core18.stepIdentity)(s)) : import_chalk15.default.dim((0, import_core18.stepIdentity)(s));
+    const label = i === plan.index ? import_chalk15.default.bold((0, import_core19.stepIdentity)(s)) : import_chalk15.default.dim((0, import_core19.stepIdentity)(s));
     console.log(`  ${marker} ${import_chalk15.default.dim(String(i))} ${label} ${import_chalk15.default.dim("(" + s.status + ")")}`);
   });
   if (plan.previousCurrentStepIdx !== runState.currentStepIdx) {
@@ -48375,6 +48539,9 @@ function printEpicDetail(epic) {
   if (epic.description) {
     console.log(import_chalk15.default.dim("  desc:     ") + epic.description);
   }
+  if (epic.tags.length) {
+    console.log(import_chalk15.default.dim("  tags:     ") + import_chalk15.default.cyan(epic.tags.join(" ")));
+  }
   if (epic.pipeline) {
     console.log(import_chalk15.default.dim("  pipeline: ") + epic.pipeline);
   }
@@ -48424,7 +48591,7 @@ function truncate(s, max) {
 
 // src/commands/recipe.ts
 var import_chalk16 = __toESM(require_source());
-var import_core19 = __toESM(require_dist());
+var import_core20 = __toESM(require_dist());
 function registerRecipe(program3) {
   const cmd = program3.command("recipe").description("Manage task-type recipes in workspace.yaml");
   cmd.command("init").description("Back-fill task-type recipes from the existing pipeline (enables --brief suggestion)").option("--dry-run", "print what would be added without writing").action((opts, actionCmd) => {
@@ -48434,7 +48601,7 @@ function registerRecipe(program3) {
       console.log(import_chalk16.default.dim(`Workspace already has ${doc.recipes.length} recipe(s) \u2014 nothing to do.`));
       return;
     }
-    const recipes = (0, import_core19.planRecipeMigration)(doc);
+    const recipes = (0, import_core20.planRecipeMigration)(doc);
     if (!recipes) {
       console.error(import_chalk16.default.yellow(
         "No recipes could be derived. A pipeline whose steps overlap the built-in\nSDLC phases (plan, design, implement, execute-test, \u2026) is required.\nApply the SDLC preset first: aidlc preset apply sdlc"
@@ -48465,7 +48632,7 @@ var path13 = __toESM(require("path"));
 var import_child_process4 = require("child_process");
 var readline = __toESM(require("readline"));
 var import_chalk17 = __toESM(require_source());
-var import_core20 = __toESM(require_dist());
+var import_core21 = __toESM(require_dist());
 var OBSERVE_PORT = 4981;
 var OBSERVE_HOST = "127.0.0.1";
 var OBSERVE_BASE = `http://${OBSERVE_HOST}:${OBSERVE_PORT}`;
@@ -48477,10 +48644,10 @@ function aidlcDataDir() {
   return path13.join(os2.homedir(), ".aidlc", "observe-data");
 }
 function settingsPath() {
-  return path13.join((0, import_core20.claudeConfigDir)(), "settings.json");
+  return path13.join((0, import_core21.claudeConfigDir)(), "settings.json");
 }
 function pluginsRoot() {
-  return path13.join((0, import_core20.claudeConfigDir)(), "plugins");
+  return path13.join((0, import_core21.claudeConfigDir)(), "plugins");
 }
 function parsePluginList(out2) {
   if (!/agents-observe/i.test(out2)) return null;
@@ -48838,14 +49005,14 @@ opening ${DASHBOARD_URL} \u2026`));
 // src/commands/ask.ts
 var import_child_process5 = require("child_process");
 var import_chalk18 = __toESM(require_source());
-var import_core21 = __toESM(require_dist());
+var import_core22 = __toESM(require_dist());
 var ASK_SYSTEM_PROMPT = `You are the AIDLC assistant, embedded in the \`aidlc\` CLI.
 Answer the user's question about AIDLC concisely and practically \u2014 prefer the
 exact command, flag, or button name over abstract description. When the user is
 getting started, give them the concrete next command to run. Use the reference
 below as ground truth and never invent commands or settings not listed in it.
 
-${import_core21.AIDLC_KNOWLEDGE}`;
+${import_core22.AIDLC_KNOWLEDGE}`;
 function registerAsk(program3) {
   program3.command("ask <prompt...>").description("ask Claude about AIDLC \u2014 setup, concepts, and CLI commands").action(async (promptParts, _opts, cmd) => {
     const userPrompt = promptParts.join(" ").trim();
@@ -48861,7 +49028,7 @@ function registerAsk(program3) {
       process.exit(1);
     }
     const root = resolveWorkspaceRoot(cmd);
-    const runner = new import_core21.DefaultRunner();
+    const runner = new import_core22.DefaultRunner();
     process.stderr.write("\n");
     const result = await runner.run({
       skill: ASK_SYSTEM_PROMPT,
@@ -48881,31 +49048,31 @@ function registerAsk(program3) {
 }
 
 // src/commands/guide.ts
-var import_core22 = __toESM(require_dist());
+var import_core23 = __toESM(require_dist());
 function registerGuide(program3) {
   program3.command("guide").description("step-by-step getting-started reference (no LLM, no cost)").action(() => {
-    console.log(import_core22.AIDLC_CLI_GUIDE_TEXT);
+    console.log(import_core23.AIDLC_CLI_GUIDE_TEXT);
   });
 }
 
 // src/commands/globals.ts
 var import_chalk19 = __toESM(require_source());
-var import_core23 = __toESM(require_dist());
+var import_core24 = __toESM(require_dist());
 function registerGlobals(program3) {
   const cmd = program3.command("globals").description("Install / uninstall built-in workflow agents + skills under ~/.claude/");
   cmd.command("status").description("Show which built-in workflows are installed globally").option("--json", "Output raw JSON").action((opts) => {
     const root = cliTemplatesRoot();
-    const rows = import_core23.BUILTIN_WORKFLOWS.map((w) => ({
+    const rows = import_core24.BUILTIN_WORKFLOWS.map((w) => ({
       id: w.id,
       name: w.name,
-      installed: (0, import_core23.isWorkflowGloballyInstalled)(root, w.id)
+      installed: (0, import_core24.isWorkflowGloballyInstalled)(root, w.id)
     }));
     if (opts.json) {
       console.log(JSON.stringify(rows, null, 2));
       return;
     }
     console.log(import_chalk19.default.bold(`
-Built-in workflow globals (${(0, import_core23.claudeConfigDir)()})`));
+Built-in workflow globals (${(0, import_core24.claudeConfigDir)()})`));
     for (const r of rows) {
       const mark = r.installed ? import_chalk19.default.green("\u2714 installed") : import_chalk19.default.dim("\xB7 not installed");
       console.log(`  ${import_chalk19.default.cyan(r.id.padEnd(24))} ${mark}`);
@@ -48914,22 +49081,22 @@ Built-in workflow globals (${(0, import_core23.claudeConfigDir)()})`));
   });
   cmd.command("install [ids...]").description("Install built-in workflow globals (default: the standard workflows)").action((ids) => {
     const root = cliTemplatesRoot();
-    const targetIds = ids.length > 0 ? ids : [...import_core23.DEFAULT_GLOBAL_WORKFLOW_IDS];
-    const known = new Set(import_core23.BUILTIN_WORKFLOWS.map((w) => w.id));
+    const targetIds = ids.length > 0 ? ids : [...import_core24.DEFAULT_GLOBAL_WORKFLOW_IDS];
+    const known = new Set(import_core24.BUILTIN_WORKFLOWS.map((w) => w.id));
     const unknown = targetIds.filter((id) => !known.has(id));
     if (unknown.length > 0) {
       console.error(import_chalk19.default.red(`Unknown workflow id(s): ${unknown.join(", ")}`));
       console.error(import_chalk19.default.dim(`Known: ${[...known].join(", ")}`));
       process.exit(1);
     }
-    const reports = (0, import_core23.installWorkflowGlobalsByIds)(root, targetIds);
+    const reports = (0, import_core24.installWorkflowGlobalsByIds)(root, targetIds);
     for (const r of reports) {
       console.log(
         import_chalk19.default.green("\u2714") + ` ${import_chalk19.default.bold(r.workflow)} \u2014 wrote ${r.written.length}, skipped ${r.skipped.length}`
       );
     }
-    console.log(import_chalk19.default.dim(`  Files live under ${(0, import_core23.claudeConfigDir)()}/{agents,skills}`));
-    const ann = (0, import_core23.installAnnotationTools)(root);
+    console.log(import_chalk19.default.dim(`  Files live under ${(0, import_core24.claudeConfigDir)()}/{agents,skills}`));
+    const ann = (0, import_core24.installAnnotationTools)(root);
     if (ann.installed) {
       console.log(
         import_chalk19.default.green("\u2714") + " annotation tooling \u2014 renderer + annotron + epic-memory + /annotate-artifact, /epic-context"
@@ -48940,8 +49107,8 @@ Built-in workflow globals (${(0, import_core23.claudeConfigDir)()})`));
   });
   cmd.command("uninstall [ids...]").description("Remove AIDLC-installed workflow globals (preserves files shared by other installed workflows)").action((ids) => {
     const root = cliTemplatesRoot();
-    const targetIds = ids.length > 0 ? ids : [...import_core23.DEFAULT_GLOBAL_WORKFLOW_IDS];
-    const reports = (0, import_core23.uninstallWorkflowGlobalsByIds)(targetIds, void 0, root);
+    const targetIds = ids.length > 0 ? ids : [...import_core24.DEFAULT_GLOBAL_WORKFLOW_IDS];
+    const reports = (0, import_core24.uninstallWorkflowGlobalsByIds)(targetIds, void 0, root);
     let removed = 0;
     for (const r of reports) {
       removed += r.removed.length;
@@ -48955,19 +49122,19 @@ Built-in workflow globals (${(0, import_core23.claudeConfigDir)()})`));
   });
   const hook = cmd.command("memory-hook <action>").description("Toggle the epic-memory auto-load hook (action: enable | disable | status)").action((action) => {
     if (action === "status") {
-      const on = (0, import_core23.isEpicMemoryHookEnabled)();
+      const on = (0, import_core24.isEpicMemoryHookEnabled)();
       console.log(`Epic-memory hook: ${on ? import_chalk19.default.green("enabled") : import_chalk19.default.dim("disabled")}`);
       return;
     }
     if (action === "enable") {
-      (0, import_core23.installAnnotationTools)(cliTemplatesRoot());
-      const r = (0, import_core23.setEpicMemoryHook)(true);
+      (0, import_core24.installAnnotationTools)(cliTemplatesRoot());
+      const r = (0, import_core24.setEpicMemoryHook)(true);
       console.log(import_chalk19.default.green("\u2714") + ` Epic-memory hook enabled${r.changed ? "" : " (already on)"}.`);
       console.log(import_chalk19.default.dim("  A prompt that mentions an epic now auto-loads its epic-memory.json."));
       return;
     }
     if (action === "disable") {
-      const r = (0, import_core23.setEpicMemoryHook)(false);
+      const r = (0, import_core24.setEpicMemoryHook)(false);
       console.log(import_chalk19.default.yellow("\u2193") + ` Epic-memory hook disabled${r.changed ? "" : " (already off)"}.`);
       return;
     }
@@ -49149,14 +49316,14 @@ function tryInstallSkill(root) {
 var fs13 = __toESM(require("fs"));
 var path15 = __toESM(require("path"));
 var import_chalk21 = __toESM(require_source());
-var import_core24 = __toESM(require_dist());
+var import_core25 = __toESM(require_dist());
 var INCIDENT_RECIPE = "native-incident";
 var FOLLOW_UP_RECIPE = "native-full";
 function registerMaintain(program3) {
   const cmd = program3.command("maintain").description("Stage 6 \u2014 turn a production signal into an incident epic (AI-Native SDLC)").option("-s, --signal <file>", 'signal JSON file ("-" reads stdin)').option("--recipe <id>", `recipe to assemble the incident pipeline from (default: ${INCIDENT_RECIPE})`).option("--pipeline <id>", "use an existing pipeline as-is instead of a recipe").option("--from <pipelineId>", "override the recipe's source pipeline").option("--epic <id>", "epic id (default: derived from the symptom, e.g. INC-CHECKOUT-RETURNS-500)").option("--prefix <prefix>", "id prefix when deriving (default: INC)").option("--json", "output the result as JSON").action((opts, actionCmd) => {
     runMaintain(opts, actionCmd);
   });
-  cmd.command("follow-up <incidentEpic>").description("Open the epic this incident calls for, with its intent.md already written").option("-s, --signal <file>", `signal JSON (default: the incident epic's ${import_core24.SIGNAL_FILE})`).option("--recipe <id>", `recipe for the new epic (default: ${FOLLOW_UP_RECIPE})`).option("--pipeline <id>", "use an existing pipeline as-is instead of a recipe").option("--from <pipelineId>", "override the recipe's source pipeline").option("--epic <id>", "id for the new epic (default: <incidentEpic>-FIX)").option("--intent <file>", "use this markdown as intent.md verbatim, instead of rendering one").option("--problem <text>", "the problem in the user's terms \u2014 no solution language").option("--who-hurts <text>", "the specific role doing the specific task").option("--cost <text>", "what it costs them (default: the signal's scope)").option("--done <text>", "the observable condition that says this cannot recur").option("--question <text>", "an open question for the human gate (repeatable)", collect, []).option("--json", "output the result as JSON").action((incidentEpic, opts, actionCmd) => {
+  cmd.command("follow-up <incidentEpic>").description("Open the epic this incident calls for, with its intent.md already written").option("-s, --signal <file>", `signal JSON (default: the incident epic's ${import_core25.SIGNAL_FILE})`).option("--recipe <id>", `recipe for the new epic (default: ${FOLLOW_UP_RECIPE})`).option("--pipeline <id>", "use an existing pipeline as-is instead of a recipe").option("--from <pipelineId>", "override the recipe's source pipeline").option("--epic <id>", "id for the new epic (default: <incidentEpic>-FIX)").option("--intent <file>", "use this markdown as intent.md verbatim, instead of rendering one").option("--problem <text>", "the problem in the user's terms \u2014 no solution language").option("--who-hurts <text>", "the specific role doing the specific task").option("--cost <text>", "what it costs them (default: the signal's scope)").option("--done <text>", "the observable condition that says this cannot recur").option("--question <text>", "an open question for the human gate (repeatable)", collect, []).option("--json", "output the result as JSON").action((incidentEpic, opts, actionCmd) => {
     runFollowUp(incidentEpic, opts, actionCmd);
   });
 }
@@ -49170,14 +49337,14 @@ function runMaintain(opts, actionCmd) {
   const root = resolveWorkspaceRoot(actionCmd);
   const doc = requireYaml(root);
   const signal = loadSignal(opts.signal);
-  const epicId = opts.epic?.trim() || (0, import_core24.followUpEpicId)(signal, { prefix: opts.prefix, taken: (0, import_core24.existingEpicIds)(root, doc) });
+  const epicId = opts.epic?.trim() || (0, import_core25.followUpEpicId)(signal, { prefix: opts.prefix, taken: (0, import_core25.existingEpicIds)(root, doc) });
   const pipelineCfg = resolvePipeline(root, doc, {
     pipeline: opts.pipeline,
     recipe: opts.recipe ?? INCIDENT_RECIPE,
     from: opts.from
   }, epicId);
   try {
-    const result = (0, import_core24.openIncidentEpic)({
+    const result = (0, import_core25.openIncidentEpic)({
       workspaceRoot: root,
       doc,
       signal,
@@ -49212,16 +49379,16 @@ function runFollowUp(incidentEpic, opts, actionCmd) {
   if (opts.signal) {
     signal = loadSignal(opts.signal);
   } else {
-    const raw = (0, import_core24.readEpicSignal)(root, doc, incidentEpic);
+    const raw = (0, import_core25.readEpicSignal)(root, doc, incidentEpic);
     if (raw === null) {
       fail2(
-        `Epic "${incidentEpic}" has no ${import_core24.SIGNAL_FILE} to open a follow-up from.`,
+        `Epic "${incidentEpic}" has no ${import_core25.SIGNAL_FILE} to open a follow-up from.`,
         `Pass one with --signal <file>, or register the incident first: aidlc maintain --signal <file>`
       );
     }
-    signal = parseOrFail(raw, path15.join(incidentEpic, import_core24.SIGNAL_FILE));
+    signal = parseOrFail(raw, path15.join(incidentEpic, import_core25.SIGNAL_FILE));
   }
-  const epicId = opts.epic?.trim() || (0, import_core24.followUpIdFor)(incidentEpic, (0, import_core24.existingEpicIds)(root, doc));
+  const epicId = opts.epic?.trim() || (0, import_core25.followUpIdFor)(incidentEpic, (0, import_core25.existingEpicIds)(root, doc));
   const pipelineCfg = resolvePipeline(root, doc, {
     pipeline: opts.pipeline,
     recipe: opts.recipe ?? FOLLOW_UP_RECIPE,
@@ -49243,7 +49410,7 @@ function runFollowUp(incidentEpic, opts, actionCmd) {
     intentMarkdown = fs13.readFileSync(opts.intent, "utf8");
   }
   try {
-    const result = (0, import_core24.openFollowUpEpic)({
+    const result = (0, import_core25.openFollowUpEpic)({
       workspaceRoot: root,
       doc,
       signal,
@@ -49282,7 +49449,7 @@ function resolvePipeline(root, doc, opts, epicId) {
   }
   let config;
   try {
-    config = (0, import_core24.validateWorkspace)(doc, ".aidlc/workspace.yaml");
+    config = (0, import_core25.validateWorkspace)(doc, ".aidlc/workspace.yaml");
   } catch (err) {
     fail2(
       "workspace.yaml is invalid \u2014 fix it before running stage 6:",
@@ -49301,26 +49468,26 @@ function resolvePipeline(root, doc, opts, epicId) {
       recipe.from = opts.from;
     }
   }
-  const pipelineId = (0, import_core24.recipePipelineId)({
+  const pipelineId = (0, import_core25.recipePipelineId)({
     recipeId: opts.recipe,
     epicId,
     taken: existingIds(doc.pipelines)
   });
   let pipelineCfg;
   try {
-    pipelineCfg = (0, import_core24.assemblePipeline)(config, { recipeId: opts.recipe, pipelineId });
+    pipelineCfg = (0, import_core25.assemblePipeline)(config, { recipeId: opts.recipe, pipelineId });
   } catch (err) {
-    if (err instanceof import_core24.PipelineAssembleError) {
+    if (err instanceof import_core25.PipelineAssembleError) {
       fail2("Could not assemble pipeline: " + err.message);
     }
     throw err;
   }
   doc.pipelines.push(pipelineCfg);
   if (epicId) {
-    (0, import_core24.stageEpicPipeline)(doc, pipelineCfg.id, epicId);
+    (0, import_core25.stageEpicPipeline)(doc, pipelineCfg.id, epicId);
   }
   try {
-    (0, import_core24.validateWorkspace)(doc, ".aidlc/workspace.yaml");
+    (0, import_core25.validateWorkspace)(doc, ".aidlc/workspace.yaml");
   } catch (err) {
     fail2(
       "Assembled pipeline failed validation \u2014 not written:",
@@ -49342,9 +49509,9 @@ function loadSignal(file) {
 }
 function parseOrFail(raw, where) {
   try {
-    return (0, import_core24.parseSignal)(raw);
+    return (0, import_core25.parseSignal)(raw);
   } catch (err) {
-    if (err instanceof import_core24.SignalParseError) {
+    if (err instanceof import_core25.SignalParseError) {
       fail2(`${where}: ${err.message}`, EXAMPLE_SIGNAL);
     }
     throw err;
@@ -49372,7 +49539,7 @@ function fail2(message, detail) {
   process.exit(1);
 }
 function rethrow(err) {
-  if (err instanceof import_core24.EpicScaffoldError) {
+  if (err instanceof import_core25.EpicScaffoldError) {
     fail2(err.message);
   }
   throw err;
@@ -49386,7 +49553,7 @@ program2.name("aidlc").description("AIDLC terminal CLI \u2014 drive workspace.ya
     setQuiet(true);
   }
   const raw = thisCommand.opts().workspace ?? process.env.AIDLC_WORKSPACE ?? process.cwd();
-  (0, import_core25.activateBackendFromWorkspace)(path16.resolve(raw), (text) => load(text));
+  (0, import_core26.activateBackendFromWorkspace)(path16.resolve(raw), (text) => load(text));
 });
 registerInit(program2);
 registerValidate(program2);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@delete101020/aidlc",
-  "version": "4.0.0",
+  "version": "4.0.2",
   "description": "AIDLC Native CLI, a fork of AIDLC — drive Claude through the six-stage AI-Native SDLC Playbook: Plan → Design → Build → Test → Deploy → Maintain, with artifact gates, requirements traceability, an incident loop back to Plan, and Spec Kit support.",
   "license": "MIT",
   "repository": {