cowork-flow 1.6.0 → 1.7.2

package/presets/claude-code/skills/cowork-flow-bootstrap/SKILL.md

@@ -0,0 +1,35 @@
+---
+name: cowork-flow-bootstrap
+description: Use only when a repository has no .cowork-flow/ runtime and the user wants to start or continue structured development work — guides installing cowork-flow into the project. Do not use when .cowork-flow/ exists; the project's own cowork-flow Skill owns that case.
+---
+
+# Cowork Flow Bootstrap
+
+This Skill is the plugin-side bootstrap. It applies to repositories that have no
+`.cowork-flow/` runtime yet. Once `.cowork-flow/` exists, stop and use the
+project's own Skills: after init, `./.cowork-flow/run task next` and the project's
+skill copies are authoritative, not this Skill.
+
+## Decide first
+
+Check the repository root for the runtime directory `.cowork-flow/` (or the
+workflow entry `./.cowork-flow/run`).
+
+## No runtime: offer the install
+
+Tell the user this repository has no cowork-flow runtime, then offer:
+
+```bash
+npx cowork-flow project init .
+```
+
+After init the project carries its own Skills (`.agents/skills/`), the task
+runtime (`.cowork-flow/`), and the workflow entry `./.cowork-flow/run task next`.
+Resume the work through that entry.
+
+## Never
+
+- Do not reproduce lifecycle rules, gates, or a second workflow document here.
+- Do not hand-create `.cowork-flow/` or copy files out of the plugin payload.
+- Do not run task lifecycle commands: without a project runtime there is no task
+  to advance.

package/presets/codex/.codex-plugin/plugin.json

@@ -0,0 +1,37 @@
+{
+  "name": "cowork-flow",
+  "version": "1.7.2",
+  "description": "AI 协作工作流引擎：任务生命周期、子代理调度、门禁与状态注入。将 cowork-flow 的完整工作流带入 Codex。",
+  "author": {
+    "name": "fishlikewater",
+    "email": "fishlikewater@126.com",
+    "url": "https://github.com/fishlikewater"
+  },
+  "homepage": "https://github.com/fishlikewater/cowork-flow",
+  "repository": "https://github.com/fishlikewater/cowork-flow",
+  "license": "MIT",
+  "keywords": [
+    "ai-agent",
+    "workflow",
+    "task-lifecycle",
+    "subagents",
+    "codex",
+    "claude-code",
+    "opencode",
+    "zcode",
+    "qoder",
+    "kimi-code",
+    "dsh"
+  ],
+  "skills": "./skills/",
+  "interface": {
+    "displayName": "Cowork Flow",
+    "shortDescription": "AI 协作工作流引擎：任务生命周期、子代理调度、门禁与状态注入。",
+    "longDescription": "cowork-flow 给 AI 编码宿主的 agent 提供可执行的工作流：任务生命周期、决策记录、规范契约与独立检查。安装后任务状态由运行时统一维护，多个宿主的会话从同一份事实起步。",
+    "developerName": "fishlikewater",
+    "category": "Developer Tools",
+    "websiteURL": "https://github.com/fishlikewater/cowork-flow",
+    "logo": "./assets/logo.svg",
+    "brandColor": "#2563EB"
+  }
+}

package/presets/codex/assets/logo.svg

@@ -0,0 +1,4 @@
+<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" width="24" height="24" role="img" aria-label="Cowork Flow">
+  <title>Cowork Flow</title>
+  <path d="M3 18H8V12H14V6H21" fill="none" stroke="#2563EB" stroke-width="3" stroke-linecap="round" stroke-linejoin="round"/>
+</svg>

package/presets/codex/skills/cowork-flow-bootstrap/SKILL.md

@@ -0,0 +1,35 @@
+---
+name: cowork-flow-bootstrap
+description: Use only when a repository has no .cowork-flow/ runtime and the user wants to start or continue structured development work — guides installing cowork-flow into the project. Do not use when .cowork-flow/ exists; the project's own cowork-flow Skill owns that case.
+---
+
+# Cowork Flow Bootstrap
+
+This Skill is the plugin-side bootstrap. It applies to repositories that have no
+`.cowork-flow/` runtime yet. Once `.cowork-flow/` exists, stop and use the
+project's own Skills: after init, `./.cowork-flow/run task next` and the project's
+skill copies are authoritative, not this Skill.
+
+## Decide first
+
+Check the repository root for the runtime directory `.cowork-flow/` (or the
+workflow entry `./.cowork-flow/run`).
+
+## No runtime: offer the install
+
+Tell the user this repository has no cowork-flow runtime, then offer:
+
+```bash
+npx cowork-flow project init .
+```
+
+After init the project carries its own Skills (`.agents/skills/`), the task
+runtime (`.cowork-flow/`), and the workflow entry `./.cowork-flow/run task next`.
+Resume the work through that entry.
+
+## Never
+
+- Do not reproduce lifecycle rules, gates, or a second workflow document here.
+- Do not hand-create `.cowork-flow/` or copy files out of the plugin payload.
+- Do not run task lifecycle commands: without a project runtime there is no task
+  to advance.

package/presets/dsh/agent.cordis.yml

@@ -1,17 +1,40 @@
 # The `cowork-flow` agent preset: the standard coding agent plus the
 # cowork-flow workflow framework (task gate, planning, fixed subagents, skills).
 #
-# This file is a copy of the deployment's `standard` preset with two changes:
-# the persona carries cowork-flow operating rules, and `skill-filesystem`
-# points at this preset's own `skills/` directory (installed alongside by
-# `cowork-flow install-dsh-preset`).
+# It started from the 0.1.x `standard` preset: the persona carries the
+# cowork-flow operating rules, and `skill-filesystem` keeps the workspace
+# default roots (`<projectRoot>/.dsh/skills` rank 100,
+# `<projectRoot>/.agents/skills` rank 200) plus the user roots, so skills come
+# from the project copy that `cwf project init` / `sync` delivers — this preset
+# ships none of its own.
 #
 # The `dsh` host adapter lives in the cowork-flow repository
 # (template/.cowork-flow/adapters/dsh/adapter.yaml) and declares this preset's
 # capability model; this file is the DSH-side composition it describes.
+#
+# Version basis: the 0.1.x line. This file is what DSH 0.1.x reads as a preset
+# directory (`$DSH_HOME/.agent-presets/<id>/`), and the 0.2.0 bundle is
+# generated from it. Each delivered form must name only the packages its own
+# line ships and carry only the config keys that line's schema accepts — either
+# mistake makes that line fail the whole preset, not just the row. The lines
+# disagree in two places: 0.1.x mounts `dsh-workflow-worker-thread`, which 0.2.0
+# dropped for `dsh-workflow-ptc`, and this `dsh-persona` row is a single
+# `text`, where 0.2.0 requires `prefix` (plus optional `suffix`). The bundle
+# generator performs both rewrites, and a third: the hook specifier becomes a
+# `file:` URL of the installed copy.
+#
+# Deliberate drift from the 0.1.x `standard` (0.1.2-rc.1), besides the
+# cowork-flow persona and the `workflow-state-hook` row: `tool-subagent` drops
+# `modelSelectionSettings`, the codex / claude-code rows use
+# `enableRunInBackground: false` where the standard takes the `backgroundMode`
+# default, `tool-web` sets `fetch: false`, and `command-goal` is not mounted.
+# Rows the 0.2.0 `standard` adds over its 0.1.x counterpart (`present`,
+# `tool-plugin-manager`) are absent from this preset on either line.
 
 # ── identity ────────────────────────────────────────────────────────────────
 
+# 0.1.x's persona schema: one `text`. 0.2.0 takes `prefix`/`suffix`, so the
+# bundle moves this wording to `prefix` unchanged (see the header).
 - id: persona
   name: '@deepseek-ai/dsh-persona'
   config:
@@ -56,16 +79,13 @@
 
 # ── skills ──────────────────────────────────────────────────────────────────
 
-# The skill REGISTRY lives in the host composition and is layered per scope:
-# these rows register into THIS preset's layer of it. `skill-filesystem`
-# contributes the cowork-flow skills shipped with this preset (`customSkillDirs`,
-# resolved against the preset's own directory via `baseUrl`) on top of the
-# local-root discovery for the current workspace.
+# The skill REGISTRY lives in the host composition and is layered per scope.
+# `skill-filesystem` discovers skills from the workspace's own roots
+# (`<projectRoot>/.dsh/skills` at rank 100, `<projectRoot>/.agents/skills` at
+# rank 200) plus the user roots, so cowork-flow skills come from the project
+# copy that `init`/`sync` delivers — the preset ships none of its own.
 - id: skill-filesystem
   name: '@deepseek-ai/dsh-skill-filesystem'
-  config:
-    customSkillDirs:
-      - !!js "process.getBuiltinModule('node:url').fileURLToPath(new URL('skills/', baseUrl))"
 
 - id: tool-skill
   name: '@deepseek-ai/dsh-tool-skill'
@@ -175,6 +195,9 @@
         enableRunInBackground: false
         maxDepth: provider-managed
 
+    # The 0.1.x row: 0.2.0 no longer ships this package and its `standard`
+    # mounts `dsh-workflow-ptc` instead, so the generated bundle swaps the row.
+    # This file itself is read by 0.1.x, where `ptc` resolves nowhere.
     - id: workflow-worker-thread
       name: '@deepseek-ai/dsh-workflow-worker-thread'
       config:

package/presets/dsh/plugins/workflow-state.js

@@ -1,25 +1,27 @@
 // cowork-flow DSH workflow-state hook.
 //
-// DSH-native equivalent of the Codex/Claude hooks
-// (template/.codex/hooks/inject-workflow-state.py and its Claude sibling):
-// this preset plugin registers ONE system-prompt section, rendered last
-// (order 1000, after identity/persona/tool guidance), whose text is the
-// `<workflow-state>` block produced by the SAME Python protocol the other
-// hosts use (adapters/host/workflow_state_hook.py), so the injected content
-// is structurally identical across hosts.
+// DSH-native equivalent of the Codex/Claude hooks:
+// registers ONE system-prompt section, rendered last
+// (order 1000, after identity/persona/tool guidance),
+// whose text is the `<workflow-state>` block produced by
+// the SAME Python protocol the other hosts use
+// (adapters/host/workflow_state_hook.py), so the injected
+// content is structurally identical across hosts.
 //
-// The block is refreshed per user message (`agent/session-start` warms it,
-// `agent/inbox/claimed` refreshes it) and after lifecycle commands settle
-// (`tools/result`), then cached per agent; each prompt assembly re-renders
-// the current cached value in place — replace semantics, no accumulation
-// across turns or steps. Between refreshes the text is byte-stable, so the
-// static prompt prefix stays cacheable and the dynamic cost is confined to
-// the trailing block.
+// Refreshed per user message (`agent/created` on 0.2.0, its former name
+// `agent/session-start` on 0.1.x, `agent/inbox/claimed` for each new one)
+// and after lifecycle commands settle (`tools/result`), then cached
+// per agent; each prompt assembly re-renders the cached
+// value in place, with replace semantics and no
+// accumulation across turns or steps. Between refreshes
+// the text is byte-stable, so the static prompt prefix
+// stays cacheable.
 //
-// Degradation is silent: no `.cowork-flow` root, a missing Python, a broken
-// runtime, or the `COWORK_FLOW_HOOKS=0` / `COWORK_FLOW_DISABLE_HOOKS=1`
-// switches all yield an empty section, and the workspace AGENTS.md gate
-// falls back to running the navigator manually.
+// Degradation is silent: no `.cowork-flow` root, a missing
+// Python, a broken runtime, or the `COWORK_FLOW_HOOKS=0` /
+// `COWORK_FLOW_DISABLE_HOOKS=1` switches all yield an
+// empty section. The workspace AGENTS.md gate then falls
+// back to running the navigator manually.
 
 import { execFile } from 'node:child_process';
 import { randomUUID } from 'node:crypto';
@@ -38,15 +40,17 @@ const EXEC_OPTIONS = {
   maxBuffer: 1024 * 1024,
 };
 const NO_ROOT_MARKER = '__COWORK_FLOW_NO_ROOT__';
-// How long a "no .cowork-flow root" verdict is trusted for one cwd. Bounded
-// on purpose: a project that installs cowork-flow mid-session becomes visible
-// after at most one TTL, at the cost of at most one interpreter probe.
+// How long a "no .cowork-flow root" verdict is trusted for
+// one cwd. Bounded on purpose: a project that installs
+// cowork-flow mid-session becomes visible after at most
+// one TTL, at the cost of at most one interpreter probe.
 const NO_ROOT_TTL_MS = 30_000;
 const noRootCache = new Map();
 
-// Python 3.9+ protocol reused from the host hook adapters. It resolves the
-// `.cowork-flow` root from the given cwd, loads the shared
-// workflow_state_hook module, and prints the full hook context block.
+// Python 3.9+ protocol reused from the host hook adapters.
+// It resolves the `.cowork-flow` root from the given cwd,
+// loads the shared workflow_state_hook module, and prints
+// the full hook context block.
 const PYTHON_PROTOCOL = `\
 import sys
 from pathlib import Path
@@ -83,9 +87,9 @@ print(build_hook_context(
 ))
 `;
 
-// The edit path sits on the finished tool result, so it gets a short leash:
-// a slow checker is dropped rather than delaying the edit the model already
-// made.
+// The edit path sits on the finished tool result, so it
+// gets a short leash: a slow checker is dropped rather
+// than delaying the edit the model already made.
 const EDIT_TIMEOUT_MS = 2500;
 const EDIT_EXEC_OPTIONS = {
   timeout: EDIT_TIMEOUT_MS,
@@ -93,9 +97,10 @@ const EDIT_EXEC_OPTIONS = {
   maxBuffer: 1024 * 1024,
 };
 
-// Editor-phase spec-check protocol: same root discovery, then the shared
-// edit-check entry so DSH sessions receive the single-line violation hint
-// the other hosts get from their PostToolUse hook. Always exits 0.
+// Editor-phase spec-check protocol: same root discovery,
+// then the shared edit-check entry so DSH sessions receive
+// the single-line violation hint the other hosts get from
+// their PostToolUse hook. Always exits 0.
 const PYTHON_EDIT_PROTOCOL = `\
 import sys
 from pathlib import Path
@@ -121,19 +126,21 @@ except Exception:
     pass
 `;
 
-// The first interpreter that ran the protocol successfully; a missing one is
-// dropped so a later refresh can rediscover (e.g. after a PATH change).
+// The first interpreter that ran the protocol
+// successfully; a missing one is dropped so a later
+// refresh can rediscover it after a PATH change.
 let workingPython = null;
 
-// Cowork-flow lifecycle commands executed through any tool. Matching the
-// command text (not the tool name) keeps the trigger stable across tool
-// surface changes.
+// Cowork-flow lifecycle commands executed through any
+// tool. Matching the command text (not the tool name)
+// keeps the trigger stable across tool surface changes.
 const LIFECYCLE_COMMAND = /\.cowork-flow[\\/]run(?:\.cmd)?(?:\s+[^\s"']*)?\s+(task|subagent|resume)\b/;
 
 
 /**
- * True when a tool argument value embeds a cowork-flow lifecycle command.
- * Accepts strings and JSON-serializable structures; never throws.
+ * True when a tool argument value embeds a cowork-flow
+ * lifecycle command. Accepts strings and JSON-serializable
+ * structures; never throws.
  */
 export function isLifecycleCommand(value) {
   try {
@@ -176,9 +183,10 @@ function tryPython(candidate, cwd, full = true) {
       EXEC_OPTIONS,
       (error, stdout) => {
         if (error) {
-          // ENOENT = the interpreter does not exist (try the next one).
-          // Any other failure means the interpreter ran but the protocol
-          // failed — the candidate works, the project runtime does not.
+          // ENOENT = the interpreter does not exist (try
+          // the next one). Any other failure means the
+          // interpreter ran but the protocol failed — the
+          // candidate works, the runtime does not.
           resolve({ missing: error.code === 'ENOENT', output: '' });
           return;
         }
@@ -198,7 +206,8 @@ function normalize(output) {
 
 
 /**
- * Test seam: forget the memoized interpreter so the next call rediscovers.
+ * Test seam: forget the memoized interpreter so the next
+ * call rediscovers.
  */
 export function resetWorkingPython() {
   workingPython = null;
@@ -206,9 +215,10 @@ export function resetWorkingPython() {
 
 
 /**
- * Nearest ancestor of `cwd` whose `.cowork-flow` entry is a directory, or
- * `null` when no such ancestor exists. Mirrors the root resolution inside
- * the Python protocol so a project without cowork-flow never pays for an
+ * Nearest ancestor of `cwd` whose `.cowork-flow` entry is
+ * a directory, or `null` when no such ancestor exists.
+ * Mirrors the root resolution inside the Python protocol,
+ * so a project without cowork-flow never pays for an
  * interpreter spawn at all.
  */
 export async function findCoworkRoot(cwd) {
@@ -232,11 +242,13 @@ export async function findCoworkRoot(cwd) {
 
 
 /**
- * Produce the `<workflow-state>` hook context block for a workspace cwd.
- * `full` selects the digest shape: true (default) for the full contract
- * block on session-start-like refreshes, false for the single fingerprint
- * line on intra-turn refreshes. Empty string means "contribute nothing"
- * (no root, hooks disabled, no usable interpreter, or a broken runtime).
+ * Produce the `<workflow-state>` hook context block for a
+ * workspace cwd. `full` selects the digest shape: true
+ * (default) for the full contract block on
+ * session-start-like refreshes, false for the single
+ * fingerprint line on intra-turn refreshes. Empty string
+ * means "contribute nothing" (no root, hooks disabled, no
+ * usable interpreter, or a broken runtime).
  */
 export async function runWorkflowState(cwd, full = true) {
   if (hooksDisabled() || !cwd) {
@@ -277,8 +289,9 @@ const EDIT_TOOL_NAMES = new Set(['write', 'edit']);
 
 
 /**
- * Path of the file a finished tool call edited, or '' for anything else.
- * DSH's fs tools expose `file_path`; other tool surfaces are ignored.
+ * Path of the file a finished tool call edited, or '' for
+ * anything else. DSH's fs tools expose `file_path`; other
+ * tool surfaces are ignored.
  */
 export function editedFilePath(exec) {
   const name = exec && typeof exec.name === 'string' ? exec.name : '';
@@ -292,9 +305,10 @@ export function editedFilePath(exec) {
 
 
 /**
- * Editor-phase spec check for one edited file. Returns the single-line
- * warning, or '' when there is nothing to say: no root, no interpreter, no
- * declaration, a clean run, or a timeout. Never throws.
+ * Editor-phase spec check for one edited file. Returns the
+ * single-line warning, or '' when there is nothing to say:
+ * no root, no interpreter, no declaration, a clean run, or
+ * a timeout. Never throws.
  */
 export async function runEditSpecCheck(cwd, filePath) {
   if (hooksDisabled() || !cwd || !filePath) {
@@ -338,9 +352,10 @@ export async function runEditSpecCheck(cwd, filePath) {
 
 
 /**
- * One identified user message carrying a plugin notice. DSH's own
- * createUserMessage lives in a package this preset deliberately does not
- * import, so the shape is built here with node's uuid.
+ * One identified user message carrying a plugin notice.
+ * DSH's own createUserMessage lives in a package this
+ * preset deliberately does not import, so the shape is
+ * built here with node's uuid.
  */
 function specNoticeMessage(text) {
   return {
@@ -358,12 +373,10 @@ function specNoticeMessage(text) {
 
 
 export function apply(ctx) {
-  // One cache entry per agent: { text, inflight, queued, full }.
-  // `full` records the digest shape the next refresh should use: session
-  // starts and inbox claims re-inject the full contract block, intra-turn
-  // refreshes after lifecycle commands only repeat the fingerprint line.
-  // The WeakMap key is the live Agent object the events and the assembly
-  // context both carry.
+  // One cache entry per agent: { text, inflight, queued,
+  // full }. `full` picks the digest shape for the next
+  // refresh: true re-injects the contract block, false
+  // repeats the fingerprint. Key: the live Agent.
   const states = new WeakMap();
 
   const refresh = (agent, full = true) => {
@@ -378,9 +391,8 @@ export function apply(ctx) {
     }
     entry.full = full;
     if (entry.inflight) {
-      // A refresh arrived while one was running (e.g. several lifecycle
-      // commands back to back). Re-run once after it settles so the latest
-      // on-disk state wins instead of being dropped.
+      // A refresh arrived mid-flight; re-run once after it
+      // settles so the latest on-disk state wins.
       entry.queued = true;
       return;
     }
@@ -401,11 +413,15 @@ export function apply(ctx) {
       });
   };
 
+  // Session-start warm-up. 0.2.0 renamed the event to `agent/created`
+  // (payload `{ agent, source, signal? }` — same `payload.agent` shape), so
+  // both names are subscribed and each DSH line keeps a warm-up path.
   ctx.on('agent/session-start', (payload) => refresh(payload && payload.agent, true));
+  ctx.on('agent/created', (payload) => refresh(payload && payload.agent, true));
   ctx.on('agent/inbox/claimed', (payload) => refresh(payload && payload.agent, true));
-  // Intra-turn refresh: a lifecycle command just settled, so the next prompt
-  // assembly should already see the new state instead of the stale block.
-  // Same-session refresh, so only the fingerprint line is repeated.
+  // Intra-turn refresh: a lifecycle command just settled, so
+  // the next assembly sees the new state, not the stale
+  // block. Only the fingerprint line is repeated.
   ctx.on('tools/result', (exec) => {
     const agent = exec && exec.agent;
     if (!agent || !isLifecycleCommand(exec && exec.arguments)) {
@@ -414,9 +430,9 @@ export function apply(ctx) {
     refresh(agent, false);
   });
 
-  // Editor-phase spec-check: the tool already ran, so this only attaches
-  // context for the next request and never blocks the call. Waterfall order
-  // requires awaiting next() before merging downstream decisions.
+  // Editor-phase spec-check: the tool already ran, so this
+  // only adds context for the next request and never blocks
+  // the call. The waterfall awaits next() first.
   ctx.on('tools/post-execute', async (exec, _result, next) => {
     const downstream = await next();
     const agent = exec && exec.agent;

package/presets/dsh/preset.yml

@@ -1,2 +1,2 @@
 name: Cowork Flow
-description: 接入 cowork-flow 流程框架的编码 Agent：任务门禁（task next）、计划绑定、runtime-context 固定子代理与 16 个流程技能（brainstorming / task-planning / task-review / agent-dispatch 等）。
+description: 接入 cowork-flow 流程框架的编码 Agent：任务门禁（task next）、计划绑定、runtime-context 固定子代理；流程技能由项目级 `.agents/skills` 提供（brainstorming / task-planning / task-review / agent-dispatch 等）。

package/presets/kimi-code/.kimi-plugin/plugin.json

@@ -0,0 +1,35 @@
+{
+  "name": "cowork-flow",
+  "version": "1.7.2",
+  "description": "AI 协作工作流引擎：任务生命周期、子代理调度、门禁与状态注入。将 cowork-flow 的完整工作流带入 Kimi Code。",
+  "author": {
+    "name": "fishlikewater",
+    "email": "fishlikewater@126.com"
+  },
+  "homepage": "https://github.com/fishlikewater/cowork-flow",
+  "license": "MIT",
+  "keywords": [
+    "ai-agent",
+    "workflow",
+    "task-lifecycle",
+    "subagents",
+    "codex",
+    "claude-code",
+    "opencode",
+    "zcode",
+    "qoder",
+    "kimi-code",
+    "dsh"
+  ],
+  "skills": "./skills/",
+  "sessionStart": {
+    "skill": "cowork-flow-bootstrap"
+  },
+  "interface": {
+    "displayName": "Cowork Flow",
+    "shortDescription": "AI 协作工作流引擎：任务生命周期、子代理调度、门禁与状态注入。",
+    "longDescription": "cowork-flow 给 AI 编码宿主的 agent 提供可执行的工作流：任务生命周期、决策记录、规范契约与独立检查。安装后任务状态由运行时统一维护，多个宿主的会话从同一份事实起步。",
+    "developerName": "fishlikewater",
+    "websiteURL": "https://github.com/fishlikewater/cowork-flow"
+  }
+}

package/presets/kimi-code/hooks/cowork-flow-inject.mjs

@@ -10,7 +10,10 @@
  *      exits 0 without spawning anything, so the user-level hook never
  *      touches unrelated repositories;
  *   2. interpreter location (COWORK_FLOW_PYTHON → python3 → python → py -3);
- *   3. stdin passthrough to inject.py, stdout/exit forwarding.
+ *   3. stdin passthrough to inject.py, stdout forwarded
+ *      and a failing project runtime absorbed (exit 0,
+ *      reason on stderr) — the hook row is machine-wide
+ *      and may meet an older project copy.
  *
  * Fail-open by construction: a missing interpreter, a missing runtime, or a
  * broken project copy all exit 0 with empty stdout instead of blocking the
@@ -103,7 +106,16 @@ function main() {
     }
     if (result.error || result.status === null) continue;
     if (result.stdout) process.stdout.write(result.stdout);
-    process.exit(result.status);
+    // The interpreter ran; its exit status is the project runtime's verdict,
+    // not ours. Forwarding it would let a stale project copy abort the submit
+    // this hook only decorates.
+    if (result.stderr) process.stderr.write(result.stderr);
+    if (result.status !== 0) {
+      process.stderr.write(
+        `cowork-flow kimi-code hook: inject.py exited ${result.status}; skipping injection\n`
+      );
+    }
+    process.exit(0);
   }
   // No interpreter could run the entry: stay silent rather than break the
   // user's submit (fail-open, same contract as the other host shims).

package/presets/kimi-code/skills/cowork-flow-bootstrap/SKILL.md

@@ -0,0 +1,35 @@
+---
+name: cowork-flow-bootstrap
+description: Use only when a repository has no .cowork-flow/ runtime and the user wants to start or continue structured development work — guides installing cowork-flow into the project. Do not use when .cowork-flow/ exists; the project's own cowork-flow Skill owns that case.
+---
+
+# Cowork Flow Bootstrap
+
+This Skill is the plugin-side bootstrap. It applies to repositories that have no
+`.cowork-flow/` runtime yet. Once `.cowork-flow/` exists, stop and use the
+project's own Skills: after init, `./.cowork-flow/run task next` and the project's
+skill copies are authoritative, not this Skill.
+
+## Decide first
+
+Check the repository root for the runtime directory `.cowork-flow/` (or the
+workflow entry `./.cowork-flow/run`).
+
+## No runtime: offer the install
+
+Tell the user this repository has no cowork-flow runtime, then offer:
+
+```bash
+npx cowork-flow project init .
+```
+
+After init the project carries its own Skills (`.agents/skills/`), the task
+runtime (`.cowork-flow/`), and the workflow entry `./.cowork-flow/run task next`.
+Resume the work through that entry.
+
+## Never
+
+- Do not reproduce lifecycle rules, gates, or a second workflow document here.
+- Do not hand-create `.cowork-flow/` or copy files out of the plugin payload.
+- Do not run task lifecycle commands: without a project runtime there is no task
+  to advance.