feat: bdt issue do

README.md

@@ -274,6 +274,17 @@ CLI flag beats `[tool.bdt.github].board` in `pyproject.toml`:
 board = "Roadmap"
 ```
 
+## `bdt issue do`
+
+```bash
+bdt issue do 60 [--agent claude] [--dry-run] [extra agent args...]
+```
+
+Hands an issue (GitHub) or work item (Azure DevOps) to a coding agent. Fetches the title and
+description and runs `claude -p "<prompt>" --name "60: <title>"`, so the session is easy to find
+with `claude --resume`. `--agent` picks another executable (it then gets just the prompt as its
+last argument); extra args are passed through; `--dry-run` prints the command.
+
 ## `bdt worktree`
 
 ```bash

bmsdna/devtools/cli.py

@@ -15,6 +15,7 @@
 from . import ado_issue, app_service_logs, commit as commit_mod
 from . import env_config
 from . import find_repo as find_repo_mod
+from . import issue_do as issue_do_mod
 from . import gh_issue, gh_pr
 from . import find_injection as find_injection_mod
 from . import lint as lint_mod
@@ -694,6 +695,31 @@ def issue_search(
         )
 
 
+@issue_app.command("do", context_settings={"allow_extra_args": True, "ignore_unknown_options": True})
+def issue_do(
+    ctx: typer.Context,
+    number: int = typer.Argument(..., help="Issue number (GitHub) or work item ID (Azure DevOps)"),
+    agent: str = typer.Option("claude", "--agent", help="Agent executable to start (claude runs headless: `claude -p ... --name '<number>: <title>'`)"),
+    dry_run: bool = typer.Option(False, "--dry-run", help="Print the agent command instead of running it"),
+    pat: str | None = typer.Option(
+        None,
+        "--pat",
+        envvar=["AZURE_DEVOPS_EXT_PAT", "AZURE_DEVOPS_PAT"],
+        help="Azure DevOps PAT (else falls back to `az` login)",
+    ),
+) -> None:
+    """Start a coding agent on an issue / work item, with the session named '<number>: <title>'.
+    Extra args after the number are passed through to the agent."""
+    remote = current_remote()
+    if isinstance(remote, GitHubRemote):
+        title, body = issue_do_mod.fetch_github(require_gh(), number)
+    else:
+        session = requests.Session()
+        session.headers.update(auth_header(pat))
+        title, body = issue_do_mod.fetch_ado(session, remote, number)
+    issue_do_mod.run(number, title, body, agent=agent, extra=list(ctx.args), dry_run=dry_run)
+
+
 @issue_app.command("update")
 def issue_update(
     number: int = typer.Argument(..., help="Issue number (GitHub) or work item ID (Azure DevOps)"),

bmsdna/devtools/issue_do.py

@@ -0,0 +1,71 @@
+"""`bdt issue do <number>`: hand an issue / work item to a coding agent.
+
+Fetches the issue's title and description (GitHub via `gh`, Azure DevOps via the REST API)
+and starts the agent headless (`claude -p`), naming the session `<number>: <title>` so it is
+easy to find again with `claude --resume`.
+"""
+
+from __future__ import annotations
+
+import json
+import re
+import shutil
+import subprocess
+import sys
+
+import requests
+
+from .ado_issue import _base_url
+from .gitrepo import AdoRemote
+
+_HTML_TAG_RE = re.compile(r"<[^>]+>")
+
+
+def fetch_github(gh: str, number: int) -> tuple[str, str]:
+    result = subprocess.run([gh, "issue", "view", str(number), "--json", "title,body"], capture_output=True, text=True, check=False)
+    if result.returncode != 0:
+        sys.exit(f"gh issue view {number} failed:\n{result.stderr.strip()}")
+    data = json.loads(result.stdout)
+    return data["title"], data.get("body") or ""
+
+
+def fetch_ado(session: requests.Session, remote: AdoRemote, number: int) -> tuple[str, str]:
+    r = session.get(
+        f"{_base_url(remote)}/_apis/wit/workitems/{number}",
+        params={"fields": "System.Title,System.Description", "api-version": "7.1"},
+    )
+    r.raise_for_status()
+    fields = r.json()["fields"]
+    # ADO descriptions are HTML; the agent only needs the text.
+    description = _HTML_TAG_RE.sub("", (fields.get("System.Description") or "").replace("<br>", "\n").replace("</div>", "\n"))
+    return fields["System.Title"], description.strip()
+
+
+def session_name(number: int, title: str) -> str:
+    return f"{number}: {title}"
+
+
+def build_prompt(number: int, title: str, body: str) -> str:
+    return (
+        f"Work on issue {number}: {title}\n\n{body.strip() or '(no description)'}\n\n"
+        "Follow the dev-workflow skill: worktree, draft PR early, implement, test, publish the PR "
+        f"and reference issue {number} in it. Do not merge."
+    )
+
+
+def build_command(agent: str, number: int, title: str, body: str, extra: list[str]) -> list[str]:
+    prompt = build_prompt(number, title, body)
+    if agent == "claude":
+        return [agent, "-p", prompt, "--name", session_name(number, title), *extra]
+    # Other agents: no known naming flag, so just hand over the prompt.
+    return [agent, *extra, prompt]
+
+
+def run(number: int, title: str, body: str, *, agent: str, extra: list[str], dry_run: bool) -> None:
+    cmd = build_command(agent, number, title, body, extra)
+    if dry_run:
+        print(" ".join(json.dumps(c) if " " in c or "\n" in c else c for c in cmd))
+        return
+    if shutil.which(agent) is None:
+        sys.exit(f"'{agent}' is required for this command but wasn't found on PATH.")
+    raise SystemExit(subprocess.run(cmd, check=False).returncode)

tests/test_issue_do.py

@@ -0,0 +1,20 @@
+from bmsdna.devtools import issue_do
+
+
+def test_claude_command_names_session_and_is_headless() -> None:
+    cmd = issue_do.build_command("claude", 60, "add command", "body text", ["--model", "opus"])
+    assert cmd[:2] == ["claude", "-p"]
+    assert "body text" in cmd[2]
+    assert cmd[3:5] == ["--name", "60: add command"]
+    assert cmd[-2:] == ["--model", "opus"]
+
+
+def test_other_agent_gets_prompt_last() -> None:
+    cmd = issue_do.build_command("codex", 7, "t", "", ["exec"])
+    assert cmd[:2] == ["codex", "exec"]
+    assert "(no description)" in cmd[2]
+
+
+def test_dry_run_prints(capsys) -> None:
+    issue_do.run(1, "x", "y", agent="claude", extra=[], dry_run=True)
+    assert '"1: x"' in capsys.readouterr().out