agex-cli 0.25.0__tar.gz → 0.26.0__tar.gz

{agex_cli-0.25.0 → agex_cli-0.26.0}/.github/workflows/publish.yml

@@ -73,8 +73,9 @@ jobs:
   # because it needs to rewrite the version before building.
   #
   # The matrix exists so we can publish the same code under three PyPI
-  # distribution names: `agex-cli` (canonical) plus the `agent-devex` and
-  # `devex-cli` aliases. Each leg rewrites `[project].name` in pyproject.toml
+  # distribution names: `devex-cli` (canonical) plus the `agent-devex` and
+  # `agex-cli` aliases (`agex-cli` is the legacy canonical name, still
+  # published for back-compat). Each leg rewrites `[project].name` in pyproject.toml
   # before building so each wheel is uploaded under its own Trusted Publishing
   # project.
   # --------------------------------------------------------------------------
@@ -97,9 +98,9 @@ jobs:
       - run: uv venv
       - run: uv pip install build
 
-      # Rewrite `[project].name` to the matrix dist name. For `agex-cli`
+      # Rewrite `[project].name` to the matrix dist name. For `devex-cli`
       # (the canonical dist) the regex still has to match so we re-emit
-      # the line; for `agent-devex` / `devex-cli` it produces an alias dist
+      # the line; for `agent-devex` / `agex-cli` it produces an alias dist
       # with identical contents. The post-rewrite re-parse is a fail-closed
       # check: if a future formatting change breaks the regex match (or
       # `[project].name` ends up wrong for any reason), the job aborts

{agex_cli-0.25.0 → agex_cli-0.26.0}/.gitignore

@@ -208,6 +208,7 @@ marimo/_lsp/
 __marimo__/
 
 # Project
+.devex/data/
 .agex/data/
 
 # Worktrees

{agex_cli-0.25.0 → agex_cli-0.26.0}/CHANGELOG.md

@@ -7,6 +7,17 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.26.0] - 2026-05-29
+
+### Added
+
+- **`devex pr await <PR> --detach` / `--check`** — non-blocking await (#64). `--detach` forks a detached background poller that runs the same readiness → CI → Sonar-gate → thread-tally logic, writes the verdict to a marker (`.devex/data/pr/<PR>/await.json`, atomic `os.replace`), and returns immediately — taking the in-session `time.sleep` (and the prompt-cache re-read tax past the ~5-min TTL) out of the agent's context. `--check` reads that marker back without sleeping or re-polling: the stored briefing + the same non-zero-on-blocked exit code when done, a "still polling (started Ns ago)" notice while running, and a clear notice (never a crash or false "clean") when no run exists or the marker schema is incompatible. The detached worker always leaves a marker — even on `gh` failure, and even if the spawn itself fails — so `--check` is never left hanging. The two flags are mutually exclusive. New per-backend footer keys (`await_detached`, `await_check_pending`, `await_check_missing`, `await_check_incompatible`) ship for all four backends.
+- **`devex pr open --detached-await`** — auto-wait at open time: after creating the PR (and posting the Qodo trigger), fork the detached `pr await` poller and return immediately, so "open → walk away → `pr await <PR> --check`" is one command. The non-blocking counterpart to `--delayed-read` (the two are mutually exclusive).
+
+### Changed
+
+- **Renamed the tool from `agex` to `devex` end to end** — `devex` is now the canonical command, `devex-cli` the canonical PyPI distribution, and the Python module is `devex` (was `agent_experience`). The on-disk state directory is `.devex/` (was `.agex/`), with a transparent read-fallback to a pre-existing `.agex/` and a one-time migration on first write. **`agex` stays a fully working, non-breaking alias**: the `agex` console script and `agex-cli` distribution are still published, `agex` is still recognized as an invoked name, and `devex explain agex` resolves to the overview page. Kept stable for back-compat: the `agex_version` config-file key, the `agex:` gamify hook-fragment IDs, and the `agex hook write` command in the shipped hook asset.
+
 ## [0.25.0] - 2026-05-29
 
 ### Added

{agex_cli-0.25.0 → agex_cli-0.26.0}/CLAUDE.md

@@ -4,24 +4,24 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## Project
 
-`agex` — non-agentic Python CLI that emits deterministic per-backend markdown briefings for autonomous agents. Published on PyPI under three distribution names: `agex-cli` (canonical), `agent-devex`, and `devex-cli` — the same wheel each time. CLI entry points: `agex` and `devex` (both invoke the same tool; emitted output reflects whichever name was typed). GitHub repo: `agentculture/devex`. Python module (unchanged): `agent_experience`.
+`devex` — non-agentic Python CLI that emits deterministic per-backend markdown briefings for autonomous agents. Published on PyPI under three distribution names: `devex-cli` (canonical), `agent-devex`, and `agex-cli` (the legacy canonical name, still published) — the same wheel each time. CLI entry points: `devex` (canonical) and `agex` (legacy alias) — both invoke the same tool; emitted output reflects whichever name was typed. GitHub repo: `agentculture/devex`. Python module: `devex` (renamed from `agent_experience`).
 
 **Source-of-truth documents:**
 
 - Spec: `docs/superpowers/specs/2026-04-18-agex-design.md`
 - Implementation plan: `docs/superpowers/plans/2026-04-18-agex-v0.1.md` — 33 tasks across 12 phases.
-- Status: PR #1 merged Phases 1–3 (scaffolding, core, `agex explain`). Phases 4–12 (`overview`, `learn`, `gamify`, `hook`, backends, docs site, tester) remain.
+- Status: PR #1 merged Phases 1–3 (scaffolding, core, `devex explain`). Phases 4–12 (`overview`, `learn`, `gamify`, `hook`, backends, docs site, tester) remain.
 
 Read the spec before any non-trivial change — the design invariants below are derived from it.
 
 ## Design invariants (non-negotiable)
 
-1. **Zero LLM calls inside agex.** All output is deterministic markdown from Jinja templates + Python.
+1. **Zero LLM calls inside devex.** All output is deterministic markdown from Jinja templates + Python.
 2. **Markdown is the only output format.** No `--json` flag.
 3. **`--agent <backend>` is required** on backend-sensitive commands. The CLI never auto-detects.
-4. **Side effects only in** `gamify`, `gamify --uninstall`, `hook write`, `pr open`, `pr reply`, `pr review`, `pr read` (journal writes), and first-run `.agex/` init. Everything else is read-only. The `agex pr` namespace allows scoped network I/O (via `gh`) and bounded `--wait` sleep — a deliberate carve-out from the no-network/no-sleep invariants.
+4. **Side effects only in** `gamify`, `gamify --uninstall`, `hook write`, `pr open`, `pr reply`, `pr review`, `pr read` (journal writes), `pr await` (journal + `--detach` marker writes under `.devex/data/pr/<pr>/` and the detached poller subprocess), and first-run `.devex/` init. Everything else is read-only. The `devex pr` namespace allows scoped network I/O (via `gh`), bounded `--wait` sleep, and — for `pr await --detach` — a detached background process that pays that sleep outside the agent session; a deliberate carve-out from the no-network/no-sleep invariants.
 5. **"Unsupported" is success** — exit 0 with a markdown notice that links to the issue tracker, not a non-zero exit.
-6. **Skills are authored by the agent, not shipped by agex.** `agex learn <topic>` teaches; `agex explain <topic>` describes; agex never writes a user skill file on the agent's behalf in v0.1.
+6. **Skills are authored by the agent, not shipped by devex.** `devex learn <topic>` teaches; `devex explain <topic>` describes; devex never writes a user skill file on the agent's behalf in v0.1.
 
 ## Architecture (3-stage pipeline)
 
@@ -36,17 +36,18 @@ cli.py ──► commands/<name>/scripts/<name>.py ──► core/render.py
                   (supported / unsupported)
 ```
 
-- `cli.py` (Typer) routes `agex <command> [args] --agent X`. No business logic.
-- Each `commands/<name>/` is a **skill-folder**: `SKILL.md` + `scripts/` + `assets/` + `references/`. The `SKILL.md` doubles as the content emitted by `agex explain <command>`.
-- `core/` is shared plumbing — backend enum, `.agex/` paths, Jinja renderer (`StrictUndefined`), TOML config, SKILL.md frontmatter parser, capability matrix loader, hook JSON I/O. Command- and content-agnostic.
+- `cli.py` (Typer) routes `devex <command> [args] --agent X`. No business logic.
+- Each `commands/<name>/` is a **skill-folder**: `SKILL.md` + `scripts/` + `assets/` + `references/`. The `SKILL.md` doubles as the content emitted by `devex explain <command>`.
+- `core/` is shared plumbing — backend enum, `.devex/` paths, Jinja renderer (`StrictUndefined`), TOML config, SKILL.md frontmatter parser, capability matrix loader, hook JSON I/O. Command- and content-agnostic.
 - A backend lives in three places: `core/backend.Backend` (enum entry), `backends/<name>/probe.py` (optional Python probe), and one YAML per relevant command under `commands/*/assets/backends/<name>.yaml`. Adding a backend touches only those locations.
 
-## `agex pr` namespace (v0.17.0+)
+## `devex pr` namespace (v0.17.0+)
 
-`lint`, `open`, `read`, `reply`, `review`, `await`, `delta`. Each command ends with a deterministic "Next step:" footer. The `pr` namespace allows scoped network I/O (via `gh`) and bounded `--wait` sleep — a deliberate carve-out from the no-network/no-sleep invariants. `pr open` (non-draft, new PR) and `pr review` post the Qodo `/agentic_review` trigger comment; the legacy `/improve` is deprecated and never emitted. The trigger string lives in one place: `commands/pr/scripts/review.QODO_REVIEW_TRIGGER`. Key modules:
+`lint`, `open`, `read`, `reply`, `review`, `await`, `delta`. Each command ends with a deterministic "Next step:" footer. The `pr` namespace allows scoped network I/O (via `gh`) and bounded `--wait` sleep — a deliberate carve-out from the no-network/no-sleep invariants. `pr open` (non-draft, new PR) and `pr review` post the Qodo `/agentic_review` trigger comment; the legacy `/improve` is deprecated and never emitted. The trigger string lives in one place: `commands/pr/scripts/review.QODO_REVIEW_TRIGGER`. `pr await --detach` / `--check` (issue #64) move the bounded poll out of the agent session: `--detach` forks a detached worker (`commands/pr/scripts/_await_worker.py`) that writes the verdict to a marker (`commands/pr/scripts/_detach.py`, atomic `os.replace`), and `--check` reads it back without sleeping. Key modules:
 
 - `core/github.py` — thin `gh` shellout wrapper; future zero-trust httpx swap touches only this file.
-- `core/journal.py` — nested-stream JSONL append/load for `.agex/data/<dir>/<stream>.jsonl`.
+- `commands/pr/scripts/_detach.py` — await-marker read/write (atomic) + the detached-subprocess spawn helper.
+- `core/journal.py` — nested-stream JSONL append/load for `.devex/data/<dir>/<stream>.jsonl`.
 - `core/backend.resolve_backend()` — `--agent` resolution with `culture.yaml` fallback.
 - `commands/pr/assets/rules/next_step_rules.py` — "Next step:" footer decision logic, per-backend phrasing under `commands/pr/assets/backends/<backend>.yaml`.
 
@@ -57,8 +58,8 @@ Spec: `docs/superpowers/specs/2026-05-10-agex-pr-design.md`. Plan: `docs/superpo
 - **Resource loading: use `importlib.resources.files(...)` and treat the result as a `Traversable`.** Call `.joinpath()` / `.is_file()` / `.read_text(encoding="utf-8")` directly. Wrap with `importlib.resources.as_file()` only when a third-party API needs a real `pathlib.Path`. Do NOT do `Path(str(files(...)))` — it's not zipapp/PEX safe.
 - **File locking: use `portalocker.lock` / `portalocker.unlock`.** Reaching for `fcntl.flock` / `msvcrt.locking` directly is a known foot-gun on Windows (see commit `923f639`).
 - **Always pass `encoding="utf-8"` to `read_text` / `write_text`.** Default locale on Windows corrupts non-ASCII output.
-- **Validate user-controlled CLI args before joining into paths.** `agex explain <topic>` rejects anything that doesn't match `^[a-z][a-z0-9-]*$` to block path traversal (commit `5ac796e`, test `test_explain_rejects_path_traversal`).
-- **Single source of truth for the version:** `pyproject.toml`. `agent_experience.__version__` derives from installed distribution metadata via `importlib.metadata.version("agex-cli")`, and `Config.agex_version` derives from `__version__` via `field(default_factory=...)`. Bumping the version means editing `pyproject.toml` only — no `__init__.py` edit needed.
+- **Validate user-controlled CLI args before joining into paths.** `devex explain <topic>` rejects anything that doesn't match `^[a-z][a-z0-9-]*$` to block path traversal (commit `5ac796e`, test `test_explain_rejects_path_traversal`).
+- **Single source of truth for the version:** `pyproject.toml`. `devex.__version__` derives from installed distribution metadata via `importlib.metadata.version("devex-cli")` (falling back to the `agent-devex` / `agex-cli` alias dists, then pyproject), and `Config.agex_version` derives from `__version__` via `field(default_factory=...)`. (The `agex_version` config-file key name is kept for back-compat with existing `.devex/config.toml` / `.agex/config.toml` files.) Bumping the version means editing `pyproject.toml` only — no `__init__.py` edit needed.
 
 ## Common commands
 
@@ -73,12 +74,12 @@ uv run pytest
 uv run pytest tests/core/test_paths.py::test_ensure_init_creates_dir_and_gitignore -v
 
 # Run the CLI
-uv run agex --version
-uv run agex explain agex
-uv run agex explain explain
+uv run devex --version
+uv run devex explain devex
+uv run devex explain explain
 
 # Coverage (matches what build.yml runs; SonarCloud reads this file)
-uv run pytest --cov=src/agent_experience --cov-report=xml --cov-report=term
+uv run pytest --cov=src/devex --cov-report=xml --cov-report=term
 ```
 
 ## CI surface
@@ -90,7 +91,7 @@ uv run pytest --cov=src/agent_experience --cov-report=xml --cov-report=term
 
 ## SonarCloud notes
 
-- `python:S5496` is **narrowly suppressed** for `src/agent_experience/core/render.py` only via `sonar.issue.ignore.multicriteria` in `sonar-project.properties`. Reason: `render_string()` renders Jinja templates that are always package-shipped; output is markdown, never HTML in a browser. See the comment in `render.py` above `render_string` for the rationale. Do not widen the suppression to other files.
+- `python:S5496` is **narrowly suppressed** for `src/devex/core/render.py` only via `sonar.issue.ignore.multicriteria` in `sonar-project.properties`. Reason: `render_string()` renders Jinja templates that are always package-shipped; output is markdown, never HTML in a browser. See the comment in `render.py` above `render_string` for the rationale. Do not widen the suppression to other files.
 - The `render.py` autoescape decision is explicit (`autoescape=select_autoescape([])`). Don't change it.
 
 ## Git workflow
@@ -100,6 +101,6 @@ uv run pytest --cov=src/agent_experience --cov-report=xml --cov-report=term
 - Push, open a PR, let CI + SonarCloud + Qodo + Copilot run. Address inline comments + resolve threads before merge.
 - Merging to `main` publishes to PyPI automatically (via `autotag` → `publish-pypi` → `github-release` in `publish.yml`). No manual tagging.
 - When posting on GitHub on the user's behalf (PR descriptions, issue replies, review-thread replies), sign so it's clear the message came from an AI. Three conventions, in priority order:
-  - **`agex pr open` / `agex pr reply`** — auto-append `- agex-cli (Claude)` (the nick is resolved from `culture.yaml` via `core.github.resolve_nick`, falling back to the repo basename). Don't sign manually.
+  - **`devex pr open` / `devex pr reply`** — auto-append `- devex-cli (Claude)` (the nick is resolved from `culture.yaml` via `core.github.resolve_nick`, falling back to the repo basename). Don't sign manually.
   - **Inside `communicate` workflow scripts** (`post-issue.sh`, `post-comment.sh`) — `agtag` resolves the same nick from `culture.yaml`. Don't sign manually.
-  - **Manual posts the scripts didn't author** (a hand-typed `gh pr create --body …`, a one-off review reply) — sign explicitly as `- agex-cli (Claude)`.
+  - **Manual posts the scripts didn't author** (a hand-typed `gh pr create --body …`, a one-off review reply) — sign explicitly as `- devex-cli (Claude)`.

{agex_cli-0.25.0 → agex_cli-0.26.0}/PKG-INFO

@@ -1,8 +1,8 @@
 Metadata-Version: 2.4
 Name: agex-cli
-Version: 0.25.0
+Version: 0.26.0
 Summary: Agent-operated developer-experience CLI — deterministic per-backend markdown briefings for autonomous agents.
-Project-URL: Homepage, https://culture.dev/agex/
+Project-URL: Homepage, https://culture.dev/devex/
 Project-URL: Repository, https://github.com/agentculture/devex
 Project-URL: Issues, https://github.com/agentculture/devex/issues
 Author: Ori Nachum
@@ -24,29 +24,29 @@ Requires-Dist: pytest>=8.0; extra == 'dev'
 Requires-Dist: syrupy>=4.6; extra == 'dev'
 Description-Content-Type: text/markdown
 
-# agex
+# devex
 
-Agent-operated developer-experience CLI. Non-agentic, deterministic, markdown-first. The same wheel is published on PyPI under three distribution names — `agex-cli` (canonical), `agent-devex`, and `devex-cli` — and installs two equivalent commands, `agex` and `devex` (emitted output reflects whichever name you invoke).
+Agent-operated developer-experience CLI. Non-agentic, deterministic, markdown-first. The same wheel is published on PyPI under three distribution names — `devex-cli` (canonical), `agent-devex`, and `agex-cli` (the legacy canonical name, still published) — and installs two equivalent commands, `devex` (canonical) and `agex` (legacy alias) (emitted output reflects whichever name you invoke).
 
 ## Install
 
 ```bash
-uv tool install agex-cli      # or: devex-cli
+uv tool install devex-cli      # or: agex-cli
 # or
-pipx install agex-cli         # or: pipx install devex-cli
+pipx install devex-cli         # or: pipx install agex-cli
 ```
 
 ## Quick start
 
 ```bash
-agex explain agex
-agex overview --agent claude-code
-agex learn --agent claude-code
+devex explain devex
+devex overview --agent claude-code
+devex learn --agent claude-code
 ```
 
 ## Docs
 
-[culture.dev/agex](https://culture.dev/agex/).
+[culture.dev/devex](https://culture.dev/devex/).
 
 Spec: `docs/superpowers/specs/2026-04-18-agex-design.md`.
 

agex_cli-0.26.0/README.md

@@ -0,0 +1,29 @@
+# devex
+
+Agent-operated developer-experience CLI. Non-agentic, deterministic, markdown-first. The same wheel is published on PyPI under three distribution names — `devex-cli` (canonical), `agent-devex`, and `agex-cli` (the legacy canonical name, still published) — and installs two equivalent commands, `devex` (canonical) and `agex` (legacy alias) (emitted output reflects whichever name you invoke).
+
+## Install
+
+```bash
+uv tool install devex-cli      # or: agex-cli
+# or
+pipx install devex-cli         # or: pipx install agex-cli
+```
+
+## Quick start
+
+```bash
+devex explain devex
+devex overview --agent claude-code
+devex learn --agent claude-code
+```
+
+## Docs
+
+[culture.dev/devex](https://culture.dev/devex/).
+
+Spec: `docs/superpowers/specs/2026-04-18-agex-design.md`.
+
+## License
+
+MIT.

{agex_cli-0.25.0 → agex_cli-0.26.0}/culture.yaml

@@ -1,3 +1,3 @@
 agents:
-- suffix: agex-cli
+- suffix: devex-cli
   backend: claude

{agex_cli-0.25.0 → agex_cli-0.26.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "agex-cli"
-version = "0.25.0"
+version = "0.26.0"
 description = "Agent-operated developer-experience CLI — deterministic per-backend markdown briefings for autonomous agents."
 authors = [{name = "Ori Nachum"}]
 license = {text = "MIT"}
@@ -26,11 +26,11 @@ dev = [
 ]
 
 [project.scripts]
-agex = "agent_experience.cli:_main_entrypoint"
-devex = "agent_experience.cli:_main_entrypoint"
+devex = "devex.cli:_main_entrypoint"
+agex = "devex.cli:_main_entrypoint"
 
 [project.urls]
-Homepage = "https://culture.dev/agex/"
+Homepage = "https://culture.dev/devex/"
 Repository = "https://github.com/agentculture/devex"
 Issues = "https://github.com/agentculture/devex/issues"
 
@@ -39,7 +39,7 @@ requires = ["hatchling"]
 build-backend = "hatchling.build"
 
 [tool.hatch.build.targets.wheel]
-packages = ["src/agent_experience"]
+packages = ["src/devex"]
 
 [tool.pytest.ini_options]
 testpaths = ["tests"]
@@ -54,7 +54,7 @@ line_length = 100
 
 [tool.bandit]
 # B311 flags any `random.*` usage as potentially unsafe for crypto purposes.
-# `core/hook_io._acquire_lock_with_retry` uses `random.uniform` purely for
+# `core/_jsonl.acquire_lock_with_retry` uses `random.uniform` purely for
 # backoff jitter between bounded retries -- not a security context, so the
 # standard `random` module is the right choice (secrets would be misuse).
 skips = ["B311"]

{agex_cli-0.25.0 → agex_cli-0.26.0}/sonar-project.properties

@@ -30,17 +30,17 @@ sonar.python.coverage.reportPaths=coverage.xml
 # S2083 (path injection) fires in commands/gamify/scripts/install.py because
 # the hooks file path is `Path.cwd() / ".claude" / "hooks.json"`. `Path.cwd()`
 # is the process's working directory (set by the user's shell before invoking
-# agex); `.claude` and `hooks.json` are hardcoded literals. No HTTP body, CLI
+# devex); `.claude` and `hooks.json` are hardcoded literals. No HTTP body, CLI
 # string argument, or other network/request input is ever joined into this
 # path — the backend name is enum-validated by `parse_backend()` before reaching
 # install.py and does not itself flow into the path. S2083 is designed for
 # web contexts where a request-body filename taints a filesystem call; that
 # threat model does not apply here. Writing `.claude/hooks.json` is the
-# documented contract of `agex gamify` (design spec invariant #4: "Side
+# documented contract of `devex gamify` (design spec invariant #4: "Side
 # effects only in gamify, gamify --uninstall, hook write, and first-run
-# `.agex/` init"). Rule is disabled only for this one file.
+# `.devex/` init"). Rule is disabled only for this one file.
 sonar.issue.ignore.multicriteria=e1,e2
 sonar.issue.ignore.multicriteria.e1.ruleKey=pythonsecurity:S5496
-sonar.issue.ignore.multicriteria.e1.resourceKey=src/agent_experience/core/render.py
+sonar.issue.ignore.multicriteria.e1.resourceKey=src/devex/core/render.py
 sonar.issue.ignore.multicriteria.e2.ruleKey=pythonsecurity:S2083
-sonar.issue.ignore.multicriteria.e2.resourceKey=src/agent_experience/commands/gamify/scripts/install.py
+sonar.issue.ignore.multicriteria.e2.resourceKey=src/devex/commands/gamify/scripts/install.py

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/__init__.py

@@ -2,14 +2,15 @@ from importlib.metadata import PackageNotFoundError, version
 
 
 def _resolve_version() -> str:
-    # `agex-cli` is the canonical PyPI distribution name; `agent-devex` and
-    # `devex-cli` are alias distributions that ship the identical wheel under
-    # different names. Whichever one is installed, surface its metadata
+    # `devex-cli` is the canonical PyPI distribution name; `agent-devex` and
+    # `agex-cli` are alias distributions that ship the identical wheel under
+    # different names (`agex-cli` is the legacy canonical name, kept published
+    # for back-compat). Whichever one is installed, surface its metadata
     # version. As a final fallback for unbuilt source checkouts (no installed
     # dist metadata), read the version directly from the repo's pyproject.toml
     # so the version stays single-sourced from pyproject.toml in every
     # reachable code path.
-    for dist in ("agex-cli", "agent-devex", "devex-cli"):
+    for dist in ("devex-cli", "agent-devex", "agex-cli"):
         try:
             return version(dist)
         except PackageNotFoundError:

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/__main__.py

@@ -1,4 +1,4 @@
-from agent_experience.cli import _main_entrypoint
+from devex.cli import _main_entrypoint
 
 if __name__ == "__main__":
     _main_entrypoint()

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/backends/acp/probe.py

@@ -1,6 +1,6 @@
 from pathlib import Path
 
-from agent_experience.backends.claude_code.probe import ProbeResult
+from devex.backends.claude_code.probe import ProbeResult
 
 
 def probe(project_dir: Path) -> ProbeResult:

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/backends/capabilities/codex.yaml

@@ -1,7 +1,7 @@
 hooks: false
 hooks_alternative: |
   Codex does not expose a hook interface. Wrap your Codex invocation in a shell
-  script that calls `agex hook write` before/after each turn as a workaround.
+  script that calls `devex hook write` before/after each turn as a workaround.
 mcp: true
 skills: false
 agents: false

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/backends/capabilities/copilot.yaml

@@ -1,7 +1,7 @@
 hooks: false
 hooks_alternative: |
   GitHub Copilot CLI does not expose a hook interface in v0.1-era versions.
-  Track usage via a shell wrapper that calls `agex hook write`.
+  Track usage via a shell wrapper that calls `devex hook write`.
 mcp: false
 skills: false
 agents: false

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/backends/claude_code/probe.py

@@ -5,7 +5,7 @@ from typing import Any
 
 import yaml
 
-from agent_experience.core.skill_loader import load_skill
+from devex.core.skill_loader import load_skill
 
 _CLAUDE_DIR = ".claude"
 

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/backends/codex/probe.py

@@ -1,6 +1,6 @@
 from pathlib import Path
 
-from agent_experience.backends.claude_code.probe import ProbeResult
+from devex.backends.claude_code.probe import ProbeResult
 
 
 def probe(project_dir: Path) -> ProbeResult:

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/backends/copilot/probe.py

@@ -1,6 +1,6 @@
 from pathlib import Path
 
-from agent_experience.backends.claude_code.probe import ProbeResult
+from devex.backends.claude_code.probe import ProbeResult
 
 
 def probe(project_dir: Path) -> ProbeResult:

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/cli.py

@@ -1,6 +1,6 @@
-"""agex CLI — stdlib argparse front end.
+"""devex CLI — stdlib argparse front end.
 
-No third-party CLI framework: this module routes `agex <command> [args]`
+No third-party CLI framework: this module routes `devex <command> [args]`
 through `argparse` only, mirroring the skeleton used by the sibling Culture
 repos (steward, devague). Business logic stays in `commands/<name>/scripts/`,
 which return ``(stdout, exit_code, stderr)`` tuples; this module just parses
@@ -13,23 +13,23 @@ import sys
 from pathlib import Path
 from typing import Optional
 
-from agent_experience import __version__
-from agent_experience.commands.doctor.scripts import doctor as doctor_script
-from agent_experience.commands.explain.scripts import explain as explain_script
-from agent_experience.commands.gamify.scripts import install as gamify_script
-from agent_experience.commands.hook.scripts import read as hook_read_script
-from agent_experience.commands.hook.scripts import write as hook_write_script
-from agent_experience.commands.learn.scripts import learn as learn_script
-from agent_experience.commands.overview.scripts import overview as overview_script
-from agent_experience.commands.pr.scripts import await_ as pr_await_script
-from agent_experience.commands.pr.scripts import delta as pr_delta_script
-from agent_experience.commands.pr.scripts import lint as pr_lint_script
-from agent_experience.commands.pr.scripts import open_ as pr_open_script
-from agent_experience.commands.pr.scripts import read as pr_read_script
-from agent_experience.commands.pr.scripts import reply as pr_reply_script
-from agent_experience.commands.pr.scripts import review as pr_review_script
-from agent_experience.core.backend import parse_backend
-from agent_experience.core.prog import prog_name
+from devex import __version__
+from devex.commands.doctor.scripts import doctor as doctor_script
+from devex.commands.explain.scripts import explain as explain_script
+from devex.commands.gamify.scripts import install as gamify_script
+from devex.commands.hook.scripts import read as hook_read_script
+from devex.commands.hook.scripts import write as hook_write_script
+from devex.commands.learn.scripts import learn as learn_script
+from devex.commands.overview.scripts import overview as overview_script
+from devex.commands.pr.scripts import await_ as pr_await_script
+from devex.commands.pr.scripts import delta as pr_delta_script
+from devex.commands.pr.scripts import lint as pr_lint_script
+from devex.commands.pr.scripts import open_ as pr_open_script
+from devex.commands.pr.scripts import read as pr_read_script
+from devex.commands.pr.scripts import reply as pr_reply_script
+from devex.commands.pr.scripts import review as pr_review_script
+from devex.core.backend import parse_backend
+from devex.core.prog import prog_name
 
 _AGENT_HELP = "Backend: claude-code, codex, copilot, or acp."
 
@@ -43,7 +43,7 @@ class _AgexArgumentParser(argparse.ArgumentParser):
     """ArgumentParser used everywhere via ``parser_class=``.
 
     argparse's native ``error()`` already prints usage to stderr and exits
-    with code 2 — which matches agex's existing bad-argument behavior — so no
+    with code 2 — which matches devex's existing bad-argument behavior — so no
     override is required. The subclass exists only so nested subparsers inherit
     it and to give a single place for any future tweak.
     """
@@ -177,6 +177,7 @@ def _cmd_pr_open(args: argparse.Namespace) -> int:
             body_file=args.body_file,
             draft=args.draft,
             delayed_read=args.delayed_read,
+            detached_await=args.detached_await,
         )
     except ValueError as exc:
         print(f"{prog_name()}: {exc}", file=sys.stderr)
@@ -224,9 +225,18 @@ def _cmd_pr_read(args: argparse.Namespace) -> int:
 
 def _cmd_pr_await(args: argparse.Namespace) -> int:
     try:
-        stdout, exit_code, stderr = pr_await_script.run(
-            agent=args.agent, project_dir=Path.cwd(), pr=args.pr, max_wait=args.max_wait
-        )
+        if args.check:
+            stdout, exit_code, stderr = pr_await_script.check(
+                agent=args.agent, project_dir=Path.cwd(), pr=args.pr
+            )
+        elif args.detach:
+            stdout, exit_code, stderr = pr_await_script.detach(
+                agent=args.agent, project_dir=Path.cwd(), pr=args.pr, max_wait=args.max_wait
+            )
+        else:
+            stdout, exit_code, stderr = pr_await_script.run(
+                agent=args.agent, project_dir=Path.cwd(), pr=args.pr, max_wait=args.max_wait
+            )
     except ValueError as exc:
         print(f"{prog_name()}: {exc}", file=sys.stderr)
         return 2
@@ -280,7 +290,7 @@ def _add_agent_option(parser: argparse.ArgumentParser, *, required: bool, help_t
 def _group_help(parser: argparse.ArgumentParser):
     """Return a handler that prints a group's help to stderr and exits 2.
 
-    Mirrors Typer's ``no_args_is_help`` for ``agex hook`` / ``agex pr`` invoked
+    Mirrors Typer's ``no_args_is_help`` for ``devex hook`` / ``devex pr`` invoked
     with no subcommand.
     """
 
@@ -305,7 +315,7 @@ def _build_parser() -> argparse.ArgumentParser:
     p_explain.set_defaults(func=_cmd_explain)
 
     # doctor
-    p_doctor = sub.add_parser("doctor", help="Diagnose the project's agex setup.")
+    p_doctor = sub.add_parser("doctor", help="Diagnose the project's devex setup.")
     p_doctor.add_argument(
         "--role", default=None, help="Render a role-specific check section (e.g., pr-review)."
     )
@@ -335,7 +345,7 @@ def _build_parser() -> argparse.ArgumentParser:
 
 
 def _register_hook(sub: argparse._SubParsersAction) -> None:
-    hook_p = sub.add_parser("hook", help="Write and read agex tracking events.")
+    hook_p = sub.add_parser("hook", help="Write and read devex tracking events.")
     hook_sub = hook_p.add_subparsers(dest="hook_command", parser_class=_AgexArgumentParser)
 
     p_write = hook_sub.add_parser("write", help="Append a tracking event.")
@@ -372,11 +382,18 @@ def _register_pr(sub: argparse._SubParsersAction) -> None:
     p_open.add_argument("--body-file", type=Path, default=None)
     p_open.add_argument("--draft", action="store_true", default=False)
     _add_agent_option(p_open, required=False, help_text=_AGENT_HELP)
-    p_open.add_argument(
+    open_wait_group = p_open.add_mutually_exclusive_group()
+    open_wait_group.add_argument(
         "--delayed-read",
         action="store_true",
         default=False,
-        help="After create, immediately run `pr read --wait 180`.",
+        help="After create, immediately run `pr read --wait 180` (blocks this session).",
+    )
+    open_wait_group.add_argument(
+        "--detached-await",
+        action="store_true",
+        default=False,
+        help="After create, fork a detached `pr await` poller; return now (read with --check).",
     )
     p_open.set_defaults(func=_cmd_pr_open)
 
@@ -417,6 +434,22 @@ def _register_pr(sub: argparse._SubParsersAction) -> None:
             "returns early (down to waited=0s) once satisfied (default 1800)."
         ),
     )
+    detach_group = p_await.add_mutually_exclusive_group()
+    detach_group.add_argument(
+        "--detach",
+        action="store_true",
+        default=False,
+        help=(
+            "Fork a background poller that writes the verdict to a marker and "
+            "return immediately (no in-session sleep). Read it later with --check."
+        ),
+    )
+    detach_group.add_argument(
+        "--check",
+        action="store_true",
+        default=False,
+        help="Read a --detach run's marker without sleeping; print the verdict (or still-polling).",
+    )
     _add_agent_option(p_await, required=False, help_text=_AGENT_HELP)
     p_await.set_defaults(func=_cmd_pr_await)
 
@@ -462,10 +495,10 @@ _KNOWN_COMMANDS = {"explain", "overview", "learn", "gamify", "hook", "doctor", "
 
 
 def _main_entrypoint() -> None:
-    """CLI entry point that routes unknown subcommands to ``agex explain agex``.
+    """CLI entry point that routes unknown subcommands to ``devex explain devex``.
 
     When the first positional argument is not a known command (and is not a
-    flag), print the ``agex explain agex`` page to stdout and the canonical
+    flag), print the ``devex explain devex`` page to stdout and the canonical
     error message to stderr, then exit with code 2.  All other invocations —
     known commands, ``--version``, ``--help``, zero-arg help — fall through to
     the normal ``main()`` dispatch unchanged.
@@ -473,9 +506,9 @@ def _main_entrypoint() -> None:
     argv = sys.argv[1:]
     if argv and not argv[0].startswith("-") and argv[0] not in _KNOWN_COMMANDS:
         print(f"{prog_name()}: error: unknown command '{argv[0]}'", file=sys.stderr)
-        # `agex` here is the explain-topic identifier (topics/agex.md), not the
+        # `devex` here is the explain-topic identifier (topics/devex.md), not the
         # invoked command name — the canonical "what is this tool" page.
-        stdout, _, _ = explain_script.run("agex")
+        stdout, _, _ = explain_script.run("devex")
         sys.stdout.write(stdout)
         sys.exit(2)
     sys.exit(main())

{agex_cli-0.25.0/src/agent_experience → agex_cli-0.26.0/src/devex}/commands/doctor/SKILL.md

@@ -1,19 +1,19 @@
 ---
 name: doctor
-description: Diagnose the agex install and the current project's `.agex/` state with a deterministic markdown health report.
+description: Diagnose the devex install and the current project's `.devex/` state with a deterministic markdown health report.
 type: command
 ---
 
-# `agex doctor`
+# `devex doctor`
 
 Run a zero-argument health check across:
 
-1. **Install** — `agex` version, Python version, package resources are reachable.
-2. **Project state** — whether `.agex/` exists in the current directory; if it does, that `config.toml` parses, `.gitignore` matches the managed content, and `data/` is writable.
+1. **Install** — `devex` version, Python version, package resources are reachable.
+2. **Project state** — whether `.devex/` exists in the current directory; if it does, that `config.toml` parses, `.gitignore` matches the managed content, and `data/` is writable.
 3. **Internal consistency** — every shipped `commands/*/SKILL.md` parses with the required frontmatter, every per-backend capability YAML loads.
-4. **Operator verification** — a short markdown checklist of things `doctor` cannot verify automatically (network reach, git-tracking of `.agex/config.toml`, agent shell-tool wiring).
+4. **Operator verification** — a short markdown checklist of things `doctor` cannot verify automatically (network reach, git-tracking of `.devex/config.toml`, agent shell-tool wiring).
 
-`doctor` is strictly read-only. It will never create `.agex/` or write anywhere on disk — if the directory is missing, that is reported as info and the command keeps going.
+`doctor` is strictly read-only. It will never create `.devex/` or write anywhere on disk — if the directory is missing, that is reported as info and the command keeps going.
 
 ## Exit codes
 
@@ -26,8 +26,8 @@ Run a zero-argument health check across:
 ## From your shell tool
 
 ```bash
-agex doctor                       # base health check
-agex doctor --role pr-review      # base + role-specific checks (when a role file ships)
+devex doctor                       # base health check
+devex doctor --role pr-review      # base + role-specific checks (when a role file ships)
 ```
 
 ## Role-specific checks
@@ -36,6 +36,6 @@ agex doctor --role pr-review      # base + role-specific checks (when a role fil
 
 ## What `doctor` does *not* do
 
-- It does not run backend probes — that's `agex overview --agent X`.
+- It does not run backend probes — that's `devex overview --agent X`.
 - It does not perform any auto-fix. Recovery instructions are emitted as markdown for the operator (agent or human) to act on.
 - It does not touch the network.