claude-teammate 0.1.22 → 0.1.24

package/README.md

@@ -102,7 +102,9 @@ npm install -g claude-teammate
 tm8 start
 ```
 
-`start` creates `.env` on the first run, then launches a background worker for the current project. If `.env` already exists with the required values, setup is skipped and your credentials are not prompted for again.
+`start` creates `~/.tm8/.env` on the first run. If that config already exists with the required values, setup is skipped and your credentials are not prompted for again.
+
+To run multiple agents on the same machine, use `--local` to store agent status and memory in the current working folder.
 
 To upgrade to latest version
 
@@ -117,32 +119,31 @@ tm8 status
 tm8 stop
 ```
 
-Per-project runtime files are stored in:
+Global runtime files and memory are stored in:
 
 ```text
-.claude-teammate/
-├── repos/
-├── worker.log
-├── worker.pid
-└── state.json
+~/.tm8/
+├── .env
+├── .claude-teammate/
+    ├── repos/
+    ├── worker.log
+    ├── worker.pid
+    └── state.json
+└── memory/
+    └── {domain}/
+        └── {workspace}/
+            ├── epic-{jira-key}.md
+            └── issue-{jira-key}.md
 ```
 
-Issue memory is stored in:
-
-```text
-memory/
-└── {domain}/
-    └── {workspace}/
-        ├── epic-{jira-key}.md
-        └── issue-{jira-key}.md
-```
+With `--local`, the same `.claude-teammate/` and `memory/` folders are created in the current working directory instead.
 
 `epic-{jira-key}.md` stores shared epic facts such as canonical epic identity, related repositories, reusable summaries, and guardrails. `issue-{jira-key}.md` stores issue-specific workflow state, clarification history, and GitHub issue tracking.
 
 The worker polls Jira once per minute. It transitions `To Do` issues to `In Progress`, asks for repo URLs when epic memory lacks them, clones all referenced repos into `.claude-teammate/repos/`, uses Claude CLI to clarify requirements from Jira comments across those repos, and creates GitHub issues once the requirements are clear enough.
 
 ```env
-# .env file generated by `tm8 start`
+# .env file generated by `tm8 start` or `tm8 start --local`
 JIRA_BASE_URL=https://yourorg.atlassian.net
 JIRA_BOT_EMAIL=bot@yourorg.com
 JIRA_BOT_API_TOKEN=...

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "claude-teammate",
-  "version": "0.1.22",
+  "version": "0.1.24",
   "description": "CLI bootstrapper for Claude Teammate.",
   "license": "MIT",
   "type": "module",

package/src/claude.js

@@ -876,7 +876,6 @@ function buildGitHubPRImplementationSystemPrompt() {
     "You are implementing a GitHub pull request plan in a local repository.",
     "You must work directly in the provided repository path and branch.",
     "If additional accessible repository paths are provided, you may read from them for source context, but only edit the primary repository unless the instructions explicitly say otherwise.",
-    "Do not use git fetch, git clone, curl, gh, or other network commands to recover missing source; rely on the provided local repository paths only.",
     "First ensure the repository is on the provided branch.",
     "Implement the plan from the pull request description.",
     "Read the epic memory snapshot before making changes so you can reuse known repo-specific build or verification knowledge and avoid repeating past mistakes.",
@@ -899,7 +898,6 @@ function buildGitHubPRCommentReviewSystemPrompt() {
   return [
     "You are triaging a human comment on an existing GitHub pull request.",
     "If additional accessible repository paths are provided, you may read from them for source context, but do not edit any repository in this step.",
-    "Do not use git fetch, git clone, curl, gh, or other network commands to recover missing source; rely on the provided local repository paths only.",
     "Read the epic memory snapshot and use it to avoid repeating known repo-specific mistakes or ignoring established constraints.",
     "You may inspect the repository and use the available tools or MCP servers if that materially helps your decision or direct reply.",
     "If the comment asks for visual verification, a snapshot, or other frontend confirmation, you may start the app, use Playwright MCP, and answer with the result directly when no code changes are needed.",