Defer sort seed checks until adjust-order startup

This commit is contained in:
2026-05-22 22:44:56 +08:00
parent e438386e68
commit 4b2eff1c7e
146 changed files with 20067 additions and 65 deletions

View File

@@ -0,0 +1,109 @@
name = "trellis-check"
description = "Workspace-write Trellis reviewer that self-fixes spec drift, lint/type-check failures, and missing tests."
sandbox_mode = "workspace-write"
developer_instructions = """
## Required: Load Trellis Context First
This platform does NOT auto-inject task context via hook. Before doing anything else, you MUST load context yourself.
### Step 1: Find the active task path
Try in order — stop at the first one that yields a task path:
1. **Look at the dispatch prompt** you received from the main agent. If its first line is `Active task: <path>` (e.g. `Active task: .trellis/tasks/04-17-foo`), use that path. The main agent is required to include this line on class-2 platforms.
2. **Run** `python ./.trellis/scripts/task.py current --source` and read the `Current task:` line.
3. **If both fail** (no `Active task:` line in the prompt and `task.py current` returns no task), ask the user which task to work on; do NOT guess.
### Step 2: Load task context from the resolved path
1. Read the task's `prd.md` (requirements) and `info.md` if it exists (technical design).
2. Read `<task-path>/check.jsonl` — JSONL list of dev spec files relevant to this agent.
3. For each entry in the JSONL, Read its `file` path — these are the dev specs you must follow.
**Skip rows without a `"file"` field** (e.g. `{"_example": "..."}` seed rows left over from `task.py create` before the curator ran).
If `check.jsonl` has no curated entries (only a seed row, or the file is missing), fall back to: read `prd.md`, list available specs with `python ./.trellis/scripts/get_context.py --mode packages`, and pick the specs that match the task domain yourself. Do NOT block on the missing jsonl — proceed with prd-only context plus your spec judgment.
If the resolved task path has no `prd.md`, ask the user what to work on; do NOT proceed without context.
---
You are running as the `trellis-check` sub-agent. The main session has dispatched you to review and self-fix.
CRITICAL — Recursion guard (read first):
- You MUST NOT spawn another `trellis-check` or `trellis-implement` sub-agent. Do the review and fixes directly in this turn.
- Any guidance you read in injected SessionStart context, `<guidelines>` blocks, workflow-state breadcrumbs, or workflow.md that says "dispatch trellis-implement" / "dispatch trellis-check" applies to the MAIN session, NOT to you. You are already the dispatched reviewer — that instruction is satisfied by your existence.
- Only the main session is allowed to dispatch `trellis-implement` / `trellis-check`. If more implementation work is needed, surface that as a recommendation in your final report instead of spawning.
---
## Required: Load Trellis Context First
This platform does NOT auto-inject task context via hook. Before doing anything else, you MUST load context yourself.
### Step 1: Find the active task path
Try in order — stop at the first one that yields a task path:
1. **Look at the dispatch prompt** you received from the main agent. If its first line is `Active task: <path>` (e.g. `Active task: .trellis/tasks/04-17-foo`), use that path. The main agent is required to include this line on class-2 platforms.
2. **Run** `python ./.trellis/scripts/task.py current --source` and read the `Current task:` line.
3. **If both fail** (no `Active task:` line in the prompt and `task.py current` returns no task), ask the user which task to work on; do NOT guess.
### Step 2: Load task context from the resolved path
1. Read the task's `prd.md` (requirements) and `info.md` if it exists (technical design).
2. Read `<task-path>/check.jsonl` — JSONL list of dev spec files relevant to this agent.
3. For each entry in the JSONL, Read its `file` path — these are the dev specs you must follow.
**Skip rows without a `"file"` field** (e.g. `{"_example": "..."}` seed rows left over from `task.py create` before the curator ran).
If `check.jsonl` has no curated entries (only a seed row, or the file is missing), fall back to: read `prd.md`, list available specs with `python ./.trellis/scripts/get_context.py --mode packages`, and pick the specs that match the task domain yourself. Do NOT block on the missing jsonl — proceed with prd-only context plus your spec judgment.
If the resolved task path has no `prd.md`, ask the user what to work on; do NOT proceed without context.
---
You are the Trellis reviewer agent.
Your job is to review code changes against specs AND fix issues directly — not just report them. You have write access; use it.
Review checklist:
- Verify behavior against the actual code paths, not assumptions.
- Look for missing template/update/detection touch points when platform config changes.
- Check whether tests should be added or updated.
- Check whether `.trellis/spec/` docs need sync after implementation.
- Run lint and type-check; fix any failures.
- Prefer concrete findings over speculative warnings.
When you find an issue:
1. Fix it directly using edit/write tools.
2. Re-run lint and type-check until green.
3. Record what you changed and why.
Output format:
## Findings (fixed)
- File: <path>
- Issue: <what was wrong>
- Fix: <what you changed>
## Findings (not fixed)
Only list issues you could not self-fix (e.g. missing product decision, out-of-scope). Explain why.
## Verification
- Lint: pass/fail
- TypeCheck: pass/fail
- Tests: pass/fail (if applicable)
If no issues are found, say so explicitly after verifying lint/type-check pass.
"""
# Disable Codex collab tools entirely for this sub-agent. With both
# multi_agent and multi_agent_v2 off, `spawn_agent` / `wait_agent` /
# `list_agents` / `close_agent` are not registered in the sub-agent's tool
# list at all — the model literally cannot call them. This is the structural
# fix for the wait_agent self-deadlock when the parent inherits its
# transcript via Codex's default `fork_turns="all"` (#240 follow-up, #241).
[features]
multi_agent = false
[features.multi_agent_v2]
enabled = false

View File

@@ -0,0 +1,90 @@
name = "trellis-implement"
description = "Workspace-write Trellis implementer that follows specs and keeps generated templates in sync."
sandbox_mode = "workspace-write"
developer_instructions = """
## Required: Load Trellis Context First
This platform does NOT auto-inject task context via hook. Before doing anything else, you MUST load context yourself.
### Step 1: Find the active task path
Try in order — stop at the first one that yields a task path:
1. **Look at the dispatch prompt** you received from the main agent. If its first line is `Active task: <path>` (e.g. `Active task: .trellis/tasks/04-17-foo`), use that path. The main agent is required to include this line on class-2 platforms.
2. **Run** `python ./.trellis/scripts/task.py current --source` and read the `Current task:` line.
3. **If both fail** (no `Active task:` line in the prompt and `task.py current` returns no task), ask the user which task to work on; do NOT guess.
### Step 2: Load task context from the resolved path
1. Read the task's `prd.md` (requirements) and `info.md` if it exists (technical design).
2. Read `<task-path>/implement.jsonl` — JSONL list of dev spec files relevant to this agent.
3. For each entry in the JSONL, Read its `file` path — these are the dev specs you must follow.
**Skip rows without a `"file"` field** (e.g. `{"_example": "..."}` seed rows left over from `task.py create` before the curator ran).
If `implement.jsonl` has no curated entries (only a seed row, or the file is missing), fall back to: read `prd.md`, list available specs with `python ./.trellis/scripts/get_context.py --mode packages`, and pick the specs that match the task domain yourself. Do NOT block on the missing jsonl — proceed with prd-only context plus your spec judgment.
If the resolved task path has no `prd.md`, ask the user what to work on; do NOT proceed without context.
---
You are running as the `trellis-implement` sub-agent. The main session has dispatched you to do the work.
CRITICAL — Recursion guard (read first):
- You MUST NOT spawn another `trellis-implement` or `trellis-check` sub-agent. Do the implementation work directly in this turn.
- Any guidance you read in injected SessionStart context, `<guidelines>` blocks, workflow-state breadcrumbs, or workflow.md that says "dispatch trellis-implement" / "dispatch trellis-check" applies to the MAIN session, NOT to you. You are already the dispatched implementer — that instruction is satisfied by your existence.
- Only the main session is allowed to dispatch `trellis-implement` / `trellis-check`. If more parallel work is needed, surface that as a recommendation in your final report instead of spawning.
---
## Required: Load Trellis Context First
This platform does NOT auto-inject task context via hook. Before doing anything else, you MUST load context yourself.
### Step 1: Find the active task path
Try in order — stop at the first one that yields a task path:
1. **Look at the dispatch prompt** you received from the main agent. If its first line is `Active task: <path>` (e.g. `Active task: .trellis/tasks/04-17-foo`), use that path. The main agent is required to include this line on class-2 platforms.
2. **Run** `python ./.trellis/scripts/task.py current --source` and read the `Current task:` line.
3. **If both fail** (no `Active task:` line in the prompt and `task.py current` returns no task), ask the user which task to work on; do NOT guess.
### Step 2: Load task context from the resolved path
1. Read the task's `prd.md` (requirements) and `info.md` if it exists (technical design).
2. Read `<task-path>/implement.jsonl` — JSONL list of dev spec files relevant to this agent.
3. For each entry in the JSONL, Read its `file` path — these are the dev specs you must follow.
**Skip rows without a `"file"` field** (e.g. `{"_example": "..."}` seed rows left over from `task.py create` before the curator ran).
If `implement.jsonl` has no curated entries (only a seed row, or the file is missing), fall back to: read `prd.md`, list available specs with `python ./.trellis/scripts/get_context.py --mode packages`, and pick the specs that match the task domain yourself. Do NOT block on the missing jsonl — proceed with prd-only context plus your spec judgment.
If the resolved task path has no `prd.md`, ask the user what to work on; do NOT proceed without context.
---
You are the Trellis implementer agent.
Rules:
- Read before write. Follow `.trellis/spec/` guidance relevant to the task.
- Keep changes focused on the requested scope.
- When touching platform registries or template lists, search first so you do not miss mirrored update paths.
- If you modify `.trellis/scripts/`, keep `packages/cli/src/templates/trellis/scripts/` in sync.
- Do not make destructive git changes unless explicitly asked.
Before finishing, summarize:
- Files changed
- Tests/checks run
- Remaining risks or follow-ups
"""
# Disable Codex collab tools entirely for this sub-agent. With both
# multi_agent and multi_agent_v2 off, `spawn_agent` / `wait_agent` /
# `list_agents` / `close_agent` are not registered in the sub-agent's tool
# list at all — the model literally cannot call them. This is the structural
# fix for the wait_agent self-deadlock when the parent inherits its
# transcript via Codex's default `fork_turns="all"` (#240 follow-up, #241).
[features]
multi_agent = false
[features.multi_agent_v2]
enabled = false

View File

@@ -0,0 +1,73 @@
name = "trellis-research"
description = "Trellis researcher for specs, code patterns, and affected files. Writes findings into {TASK_DIR}/research/ — read-only elsewhere."
sandbox_mode = "workspace-write"
developer_instructions = """
You are the Trellis researcher agent.
## Core principle
Conversations get compacted; files don't. Every research topic MUST be
persisted to `{TASK_DIR}/research/<topic>.md`. Returning findings only
through the chat reply is a failure.
## Workflow
1. Run `python ./.trellis/scripts/task.py current --source` to get the
active task path and source. If no active task is set, ask the user
where to write output; do not guess.
2. Run `mkdir -p <TASK_DIR>/research` to ensure the directory exists.
3. Read `.trellis/workflow.md`, relevant `.trellis/spec/` files, and
target code before forming an opinion.
4. For each research topic, write `<TASK_DIR>/research/<slug>.md` with:
- Query, scope, date
- Files found (path + one-line description)
- Code patterns (cite file:line)
- External references (docs, versions)
- Related specs
- Caveats / not-found notes
5. Reply with only: list of files written, one-line summary per file,
any critical caveats. Do not paste full research into the reply.
## Scope limits
Write allowed ONLY in `{TASK_DIR}/research/`.
Write forbidden everywhere else:
- Code files (`src/`, `lib/`, …)
- Spec files (`.trellis/spec/`) — use `update-spec` skill instead
- `.trellis/scripts/`, `.trellis/workflow.md`, platform config
- Other task directories
- Any git operation
If the user asks you to edit code, decline and tell them to spawn the
`implement` agent.
## Output format for each research file
```
# Research: <topic>
- Query: ...
- Scope: internal / external / mixed
- Date: YYYY-MM-DD
## Findings
...
## Caveats / Not Found
...
```
"""
# Disable Codex collab tools entirely for this sub-agent. With both
# multi_agent and multi_agent_v2 off, `spawn_agent` / `wait_agent` /
# `list_agents` / `close_agent` are not registered in the sub-agent's tool
# list at all — the model literally cannot call them. This is the structural
# fix for the wait_agent self-deadlock when the parent inherits its
# transcript via Codex's default `fork_turns="all"` (#240 follow-up, #241).
[features]
multi_agent = false
[features.multi_agent_v2]
enabled = false

28
.codex/config.toml Normal file
View File

@@ -0,0 +1,28 @@
# Project-scoped Codex defaults for Trellis workflows.
# Codex merges this layer after the user-level config when the project
# is marked as a trusted project. To trust this project, add it under
# `[projects]` in ~/.codex/config.toml, e.g.:
#
# [projects."/abs/path/to/this/repo"]
# trust_level = "trusted"
# Keep AGENTS.md as the primary project instruction file.
project_doc_fallback_filenames = ["AGENTS.md"]
# Codex hooks (`hooks.json` in this directory) only fire when the user
# has enabled them in their USER-level config: `[features].hooks = true`
# in ~/.codex/config.toml (Codex 0.129+; legacy name: `codex_hooks = true`,
# still works but emits a deprecation warning on 0.129+). Project-level
# config.toml cannot set feature flags; they must be user-level.
# Codex 0.129+ additionally gates each installed hook behind a one-time
# `/hooks` TUI review; until the user approves it, the hook stays inactive.
# NOTE: Trellis intentionally does NOT write a [features.multi_agent_v2]
# block here. Codex CLI changed `features` deserialization between 0.130
# and 0.131: the structured table form (with max_concurrent_threads_per_session
# / *_wait_timeout_ms) is only accepted by 0.131+. On 0.130 and earlier —
# including the codex CLI bundled inside the Codex desktop app — it fails
# with `data did not match any variant of untagged enum FeatureToml`, which
# aborts the entire config load and blocks Codex from starting. Codex's own
# default for multi_agent_v2 is used instead; tune it in your user-level
# config if needed.

15
.codex/hooks.json Normal file
View File

@@ -0,0 +1,15 @@
{
"hooks": {
"UserPromptSubmit": [
{
"hooks": [
{
"type": "command",
"command": "python -X utf8 .codex/hooks/inject-workflow-state.py",
"timeout": 15
}
]
}
]
}
}

View File

@@ -0,0 +1,387 @@
#!/usr/bin/env python3
"""Trellis per-turn breadcrumb hook (UserPromptSubmit / BeforeAgent equivalent).
Runs on every user prompt. Resolves the active task through Trellis'
session-aware active task resolver and emits a short <workflow-state>
block reminding the main AI what task is active and its expected flow.
The emitted ``hookEventName`` field is platform-aware: most hosts expect
``UserPromptSubmit`` (Claude Code naming, also accepted by Cursor / Qoder /
CodeBuddy / Droid / Codex / Copilot wiring), but Gemini CLI 0.40.x renamed
its per-turn event to ``BeforeAgent`` and its schema validator rejects the
legacy name. ``_detect_platform`` picks the right value at runtime.
Breadcrumb text is pulled exclusively from workflow.md
[workflow-state:STATUS] tag blocks — workflow.md is the single source of
truth. There are no fallback dicts in this script: when workflow.md is
missing or a tag is absent, the breadcrumb degrades to a generic
"Refer to workflow.md for current step." line so users see (and fix)
the broken state instead of the hook silently masking it.
Shared across all hook-capable platforms (Claude, Cursor, Codex, Qoder,
CodeBuddy, Droid, Gemini, Copilot). Kiro is not wired (no per-turn
hook entry point). Written to each platform's hooks directory via
writeSharedHooks() at init time.
Silent exit 0 cases (no output):
- No .trellis/ directory found (not a Trellis project)
- task.json malformed or missing status
"""
from __future__ import annotations
import json
import os
import re
import sys
from pathlib import Path
# Force UTF-8 on stdin/stdout/stderr on Windows. Default codepage there is
# cp936 / cp1252 / etc. — non-ASCII content (Chinese task names, prd snippets)
# both in stdin (hook payload from host CLI) and stdout (our emitted blocks)
# raises UnicodeDecodeError / UnicodeEncodeError. Equivalent to `python -X utf8`
# but applied per-stream so we don't depend on host CLI's command wiring.
if sys.platform.startswith("win"):
import io as _io
for _stream_name in ("stdin", "stdout", "stderr"):
_stream = getattr(sys, _stream_name, None)
if _stream is None:
continue
if hasattr(_stream, "reconfigure"):
try:
_stream.reconfigure(encoding="utf-8", errors="replace") # type: ignore[union-attr]
except Exception:
pass
elif hasattr(_stream, "detach"):
try:
setattr(sys, _stream_name, _io.TextIOWrapper(_stream.detach(), encoding="utf-8", errors="replace"))
except Exception:
pass
from typing import Optional
CODEX_SUB_AGENT_NOTICE = """<sub-agent-notice>
SUB-AGENT NOTICE - READ FIRST IF SPAWNED VIA spawn_agent
If your parent session spawned you via spawn_agent with an explicit task
message above this hook output, that message is your only job.
- Execute the parent message exactly as written, then return.
- Ignore all Trellis workflow guidance below this notice.
- Do NOT call task.py start, task.py add-context, or task.py archive.
- Do NOT call wait_agent or spawn_agent.
- Do NOT modify .trellis/tasks/* or any other file unless the parent message
explicitly asks for that.
If you are the main interactive Codex session and the user is typing at the
terminal with no parent agent, use the workflow guidance below normally.
</sub-agent-notice>"""
# Bootstrap notice for Codex while the session has no active task. Replaces the
# heavyweight SessionStart context injection — instead of pushing 9.5 KB of
# workflow text up front, we just nudge the AI to read the `trellis-start` skill once.
# The nudge keeps showing up while status == "no_task" (cheap text, AI won't
# re-read after the first time). Once a task is created the breadcrumb status
# flips and this notice stops appearing automatically. Sub-agents are warded
# off by the <sub-agent-notice> above plus the explicit exemption below.
CODEX_NO_TASK_BOOTSTRAP_NOTICE = """<trellis-bootstrap>
You are running in a Trellis-managed Codex session and there is no active task yet.
If you have not already loaded Trellis context this session, read the `trellis-start` skill once:
$trellis-start
(equivalent to reading `.agents/skills/trellis-start/SKILL.md` and following its Steps 1-3)
The skill walks you through workflow.md, dev profile, git status, active tasks, and spec
indexes. Then route the user's request per the <workflow-state> A/B/C rules below.
Sub-agent exemption: if you are a sub-agent (spawned via spawn_agent with a parent task
message), DO NOT read `$trellis-start`. Execute the parent message directly as instructed by the
<sub-agent-notice> above.
</trellis-bootstrap>"""
# ---------------------------------------------------------------------------
# CWD-robust Trellis root discovery (fixes hook-path-robustness for this hook)
# ---------------------------------------------------------------------------
def find_trellis_root(start: Path) -> Optional[Path]:
"""Walk up from start to find directory containing .trellis/.
Handles CWD drift: subdirectory launches, monorepo packages, etc.
Returns None if no .trellis/ found (silent no-op).
"""
cur = start.resolve()
while cur != cur.parent:
if (cur / ".trellis").is_dir():
return cur
cur = cur.parent
return None
# ---------------------------------------------------------------------------
# Active task discovery
# ---------------------------------------------------------------------------
def _detect_platform(input_data: dict) -> str | None:
if isinstance(input_data.get("cursor_version"), str):
return "cursor"
env_map = {
"CLAUDE_PROJECT_DIR": "claude",
"CURSOR_PROJECT_DIR": "cursor",
"CODEBUDDY_PROJECT_DIR": "codebuddy",
"FACTORY_PROJECT_DIR": "droid",
"GEMINI_PROJECT_DIR": "gemini",
"QODER_PROJECT_DIR": "qoder",
"KIRO_PROJECT_DIR": "kiro",
"COPILOT_PROJECT_DIR": "copilot",
}
for env_name, platform in env_map.items():
if os.environ.get(env_name):
return platform
script_parts = set(Path(sys.argv[0]).parts)
if ".claude" in script_parts:
return "claude"
if ".cursor" in script_parts:
return "cursor"
if ".codex" in script_parts:
return "codex"
if ".gemini" in script_parts:
return "gemini"
if ".qoder" in script_parts:
return "qoder"
if ".codebuddy" in script_parts:
return "codebuddy"
if ".factory" in script_parts:
return "droid"
if ".kiro" in script_parts:
return "kiro"
return None
def _resolve_active_task(root: Path, input_data: dict):
scripts_dir = root / ".trellis" / "scripts"
if str(scripts_dir) not in sys.path:
sys.path.insert(0, str(scripts_dir))
from common.active_task import resolve_active_task # type: ignore[import-not-found]
return resolve_active_task(root, input_data, platform=_detect_platform(input_data))
def get_active_task(root: Path, input_data: dict) -> Optional[tuple[str, str, str]]:
"""Return (task_id, status, source) from the current active task."""
active = _resolve_active_task(root, input_data)
if not active.task_path:
return None
task_dir = Path(active.task_path)
if not task_dir.is_absolute():
task_dir = root / task_dir
if active.stale:
return task_dir.name, f"stale_{active.source_type}", active.source
task_json = task_dir / "task.json"
if not task_json.is_file():
return None
try:
data = json.loads(task_json.read_text(encoding="utf-8"))
except (json.JSONDecodeError, OSError):
return None
task_id = data.get("id") or task_dir.name
status = data.get("status", "")
if not isinstance(status, str) or not status:
return None
return task_id, status, active.source
# ---------------------------------------------------------------------------
# Breadcrumb loading: parse workflow.md, fall back to hardcoded defaults
# ---------------------------------------------------------------------------
# Supports STATUS values with letters, digits, underscores, hyphens
# (so "in-review" / "blocked-by-team" work alongside "in_progress").
_TAG_RE = re.compile(
r"\[workflow-state:([A-Za-z0-9_-]+)\]\s*\n(.*?)\n\s*\[/workflow-state:\1\]",
re.DOTALL,
)
def load_breadcrumbs(root: Path) -> dict[str, str]:
"""Parse workflow.md for [workflow-state:STATUS] blocks.
Returns {status: body_text}. workflow.md is the single source of
truth — there are no fallback dicts in this script. Missing tags
(or a missing/unreadable workflow.md) fall back to a generic line
in build_breadcrumb so users see the broken state and fix
workflow.md, rather than the hook silently masking the issue.
"""
workflow = root / ".trellis" / "workflow.md"
if not workflow.is_file():
return {}
try:
content = workflow.read_text(encoding="utf-8")
except OSError:
return {}
result: dict[str, str] = {}
for match in _TAG_RE.finditer(content):
status = match.group(1)
body = match.group(2).strip()
if body:
result[status] = body
return result
def _read_trellis_config(root: Path) -> dict:
"""Load .trellis/config.yaml via the bundled trellis_config helper.
The helper lives in .trellis/scripts/common; the hook lives outside the
scripts tree, so we extend sys.path before importing.
"""
scripts_dir = root / ".trellis" / "scripts"
if str(scripts_dir) not in sys.path:
sys.path.insert(0, str(scripts_dir))
try:
from common.trellis_config import read_trellis_config # type: ignore[import-not-found]
except Exception:
return {}
try:
return read_trellis_config(root)
except Exception:
return {}
def _codex_mode_banner(config: dict) -> str:
"""Emit a `<codex-mode>` banner for the additionalContext payload.
Reads `codex.dispatch_mode` from .trellis/config.yaml; defaults to
`inline` when missing or invalid because Codex sub-agents run with
`fork_turns="none"` isolation and can't inherit the parent session's
task context. The banner makes the active mode explicit to Codex AI
per turn, complementing the workflow-state body which is per-status.
Mode tells AI which dispatch protocol to follow; workflow-state tells
AI what step it's at.
"""
mode = "inline"
if isinstance(config, dict):
codex_cfg = config.get("codex")
if isinstance(codex_cfg, dict):
cfg_mode = codex_cfg.get("dispatch_mode")
if cfg_mode in ("inline", "sub-agent"):
mode = cfg_mode
return f"<codex-mode>{mode}</codex-mode>"
def resolve_breadcrumb_key(
status: str, platform: str | None, config: dict
) -> str:
"""Pick the breadcrumb tag key based on Codex dispatch_mode.
Codex defaults to ``inline`` because sub-agents run with ``fork_turns="none"``
isolation and can't inherit the parent session's task context. Users can
opt into ``codex.dispatch_mode: sub-agent`` in ``.trellis/config.yaml``
to use the parallel ``<status>-inline`` tag → ``<status>`` flip. Invalid
or missing values fall back to inline.
Non-codex platforms return the plain status unchanged.
"""
if platform == "codex":
mode = "inline"
if isinstance(config, dict):
codex_cfg = config.get("codex")
if isinstance(codex_cfg, dict):
cfg_mode = codex_cfg.get("dispatch_mode")
if cfg_mode in ("inline", "sub-agent"):
mode = cfg_mode
return f"{status}-inline" if mode == "inline" else status
return status
def build_breadcrumb(
task_id: Optional[str],
status: str,
templates: dict[str, str],
source: str | None = None,
breadcrumb_key: str | None = None,
) -> str:
"""Build the <workflow-state>...</workflow-state> block.
- Known status (tag present in workflow.md) → detailed template body
- Unknown status (no tag, or workflow.md missing) → generic
"Refer to workflow.md for current step." line
- `no_task` pseudo-status (task_id is None) → header omits task info
"""
lookup_key = breadcrumb_key or status
body = templates.get(lookup_key)
if body is None and lookup_key != status:
body = templates.get(status)
if body is None:
body = "Refer to workflow.md for current step."
header = f"Status: {status}" if task_id is None else f"Task: {task_id} ({status})"
if source:
header = f"{header}\nSource: {source}"
return f"<workflow-state>\n{header}\n{body}\n</workflow-state>"
# ---------------------------------------------------------------------------
# Entry
# ---------------------------------------------------------------------------
def main() -> int:
if os.environ.get("TRELLIS_HOOKS") == "0" or os.environ.get("TRELLIS_DISABLE_HOOKS") == "1":
return 0
try:
data = json.load(sys.stdin)
except (json.JSONDecodeError, ValueError):
data = {}
cwd_str = data.get("cwd") or os.getcwd()
cwd = Path(cwd_str)
root = find_trellis_root(cwd)
if root is None:
return 0 # not a Trellis project
templates = load_breadcrumbs(root)
platform = _detect_platform(data)
config = _read_trellis_config(root)
task = get_active_task(root, data)
if task is None:
# No active task — still emit a breadcrumb nudging AI toward
# trellis-brainstorm + task.py create when user describes real work.
no_task_key = resolve_breadcrumb_key("no_task", platform, config)
breadcrumb = build_breadcrumb(
None, "no_task", templates, breadcrumb_key=no_task_key
)
else:
task_id, status, source = task
status_key = resolve_breadcrumb_key(status, platform, config)
breadcrumb = build_breadcrumb(
task_id, status, templates, source, breadcrumb_key=status_key
)
if platform == "codex":
parts: list[str] = [CODEX_SUB_AGENT_NOTICE]
if task is None:
parts.append(CODEX_NO_TASK_BOOTSTRAP_NOTICE)
parts.append(_codex_mode_banner(config))
parts.append(breadcrumb)
breadcrumb = "\n\n".join(parts)
# Gemini CLI 0.40.x rejects "UserPromptSubmit" — its per-turn event is
# named "BeforeAgent". Other platforms (Claude/Cursor/Qoder/CodeBuddy/
# Droid/Codex/Copilot) accept the original Claude-style name.
hook_event_name = (
"BeforeAgent" if platform == "gemini" else "UserPromptSubmit"
)
output = {
"hookSpecificOutput": {
"hookEventName": hook_event_name,
"additionalContext": breadcrumb,
}
}
print(json.dumps(output))
return 0
if __name__ == "__main__":
sys.exit(main())

View File

@@ -0,0 +1,481 @@
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Codex Session Start Hook - Inject Trellis context into Codex sessions.
Output format follows Codex hook protocol:
stdout JSON → { hookSpecificOutput: { hookEventName: "SessionStart", additionalContext: "..." } }
"""
from __future__ import annotations
import json
import os
import re
import subprocess
import sys
import warnings
from io import StringIO
from pathlib import Path
# Force UTF-8 on stdin/stdout/stderr on Windows. Default codepage there is
# cp936 / cp1252 / etc. — non-ASCII content (Chinese task names, prd snippets)
# both in stdin (hook payload from host CLI) and stdout (our emitted blocks)
# raises UnicodeDecodeError / UnicodeEncodeError. Equivalent to `python -X utf8`
# but applied per-stream so we don't depend on host CLI's command wiring.
if sys.platform.startswith("win"):
import io as _io
for _stream_name in ("stdin", "stdout", "stderr"):
_stream = getattr(sys, _stream_name, None)
if _stream is None:
continue
if hasattr(_stream, "reconfigure"):
try:
_stream.reconfigure(encoding="utf-8", errors="replace") # type: ignore[union-attr]
except Exception:
pass
elif hasattr(_stream, "detach"):
try:
setattr(sys, _stream_name, _io.TextIOWrapper(_stream.detach(), encoding="utf-8", errors="replace"))
except Exception:
pass
def _normalize_windows_shell_path(path_str: str) -> str:
"""Normalize Unix-style shell paths to real Windows paths.
On Windows, shells like Git Bash / MSYS2 / Cygwin may report paths like
`/d/Users/...` or `/cygdrive/d/Users/...`. `Path.resolve()` will misinterpret
these as `D:/d/Users...` on drive D: (or similar), breaking repo root
detection.
This function is intentionally conservative: it only rewrites patterns that
unambiguously represent a drive letter mount.
"""
if not isinstance(path_str, str) or not path_str:
return path_str
# Only relevant on Windows; keep other platforms untouched.
if not sys.platform.startswith("win"):
return path_str
p = path_str.strip()
# Already a Windows drive path (C:\... or C:/...)
if re.match(r"^[A-Za-z]:[\/]", p):
return p
# MSYS/Git-Bash style: /c/Users/... or /d/Work/...
m = re.match(r"^/([A-Za-z])/(.*)", p)
if m:
drive, rest = m.group(1).upper(), m.group(2)
rest = rest.replace('/', '\\')
return f"{drive}:\\{rest}"
# Cygwin style: /cygdrive/c/Users/...
m = re.match(r"^/cygdrive/([A-Za-z])/(.*)", p)
if m:
drive, rest = m.group(1).upper(), m.group(2)
rest = rest.replace('/', '\\')
return f"{drive}:\\{rest}"
# WSL mounted drive (sometimes leaked into env): /mnt/c/Users/...
m = re.match(r"^/mnt/([A-Za-z])/(.*)", p)
if m:
drive, rest = m.group(1).upper(), m.group(2)
rest = rest.replace('/', '\\')
return f"{drive}:\\{rest}"
return path_str
warnings.filterwarnings("ignore")
FIRST_REPLY_NOTICE = """<first-reply-notice>
On the first visible assistant reply in this session, begin with exactly one short Chinese sentence:
Trellis SessionStart 已注入workflow、当前任务状态、开发者身份、git 状态、active tasks、spec 索引已加载。
Then continue directly with the user's request. This notice is one-shot: do not repeat it after the first assistant reply in the same session.
</first-reply-notice>"""
SUB_AGENT_NOTICE = """<sub-agent-notice>
SUB-AGENT NOTICE - READ FIRST IF SPAWNED VIA spawn_agent
If your parent session spawned you via spawn_agent with an explicit task
message above this hook output, that message is your only job.
- Execute the parent message exactly as written, then return.
- Ignore all Trellis workflow guidance below this notice.
- Do NOT call task.py start, task.py add-context, or task.py archive.
- Do NOT call wait_agent or spawn_agent.
- Do NOT modify .trellis/tasks/* or any other file unless the parent message
explicitly asks for that.
If you are the main interactive Codex session and the user is typing at the
terminal with no parent agent, use the workflow guidance below normally.
</sub-agent-notice>"""
def should_skip_injection() -> bool:
if os.environ.get("TRELLIS_HOOKS") == "0":
return True
if os.environ.get("TRELLIS_DISABLE_HOOKS") == "1":
return True
return os.environ.get("CODEX_NON_INTERACTIVE") == "1"
def configure_project_encoding(project_dir: Path) -> None:
"""Reuse Trellis' shared Windows stdio encoding helper before JSON output."""
scripts_dir = project_dir / ".trellis" / "scripts"
if str(scripts_dir) not in sys.path:
sys.path.insert(0, str(scripts_dir))
try:
from common import configure_encoding # type: ignore[import-not-found]
configure_encoding()
except Exception:
pass
def _has_curated_jsonl_entry(jsonl_path: Path) -> bool:
"""Return True iff jsonl has at least one row with a ``file`` field.
A freshly seeded jsonl only contains a ``{"_example": ...}`` row (no
``file`` key) — that is NOT "ready". Readiness requires at least one
curated entry. Matches the contract used by ``inject-subagent-context.py``.
"""
try:
for line in jsonl_path.read_text(encoding="utf-8").splitlines():
line = line.strip()
if not line:
continue
try:
row = json.loads(line)
except json.JSONDecodeError:
continue
if isinstance(row, dict) and row.get("file"):
return True
except (OSError, UnicodeDecodeError):
return False
return False
def read_file(path: Path, fallback: str = "") -> str:
try:
return path.read_text(encoding="utf-8")
except (FileNotFoundError, PermissionError):
return fallback
def _resolve_context_key(project_dir: Path, hook_input: dict) -> str | None:
scripts_dir = project_dir / ".trellis" / "scripts"
if str(scripts_dir) not in sys.path:
sys.path.insert(0, str(scripts_dir))
try:
from common.active_task import resolve_context_key # type: ignore[import-not-found]
except Exception:
return None
return resolve_context_key(hook_input, platform="codex")
def _resolve_active_task(trellis_dir: Path, hook_input: dict):
scripts_dir = trellis_dir / "scripts"
if str(scripts_dir) not in sys.path:
sys.path.insert(0, str(scripts_dir))
from common.active_task import resolve_active_task # type: ignore[import-not-found]
return resolve_active_task(trellis_dir.parent, hook_input, platform="codex")
def run_script(script_path: Path, context_key: str | None = None) -> str:
try:
env = os.environ.copy()
env["PYTHONIOENCODING"] = "utf-8"
if context_key:
env["TRELLIS_CONTEXT_ID"] = context_key
cmd = [sys.executable, "-W", "ignore", str(script_path)]
result = subprocess.run(
cmd,
capture_output=True,
text=True,
encoding="utf-8",
errors="replace",
timeout=5,
cwd=str(script_path.parent.parent.parent),
env=env,
)
return result.stdout if result.returncode == 0 else "No context available"
except (subprocess.TimeoutExpired, FileNotFoundError, PermissionError):
return "No context available"
def _normalize_task_ref(task_ref: str) -> str:
normalized = task_ref.strip()
if not normalized:
return ""
path_obj = Path(normalized)
if path_obj.is_absolute():
return str(path_obj)
normalized = normalized.replace("\\", "/")
while normalized.startswith("./"):
normalized = normalized[2:]
if normalized.startswith("tasks/"):
return f".trellis/{normalized}"
return normalized
def _resolve_task_dir(trellis_dir: Path, task_ref: str) -> Path:
normalized = _normalize_task_ref(task_ref)
path_obj = Path(normalized)
if path_obj.is_absolute():
return path_obj
if normalized.startswith(".trellis/"):
return trellis_dir.parent / path_obj
return trellis_dir / "tasks" / path_obj
def _get_task_status(trellis_dir: Path, hook_input: dict) -> str:
active = _resolve_active_task(trellis_dir, hook_input)
if not active.task_path:
return f"Status: NO ACTIVE TASK\nSource: {active.source}\nNext: Describe what you want to work on"
task_ref = active.task_path
task_dir = _resolve_task_dir(trellis_dir, task_ref)
if active.stale or not task_dir.is_dir():
return f"Status: STALE POINTER\nTask: {task_ref}\nSource: {active.source}\nNext: Task directory not found. Run: python ./.trellis/scripts/task.py finish"
task_json_path = task_dir / "task.json"
task_data: dict = {}
if task_json_path.is_file():
try:
task_data = json.loads(task_json_path.read_text(encoding="utf-8"))
except (json.JSONDecodeError, PermissionError):
pass
task_title = task_data.get("title", task_ref)
task_status = task_data.get("status", "unknown")
if task_status == "completed":
return f"Status: COMPLETED\nTask: {task_title}\nSource: {active.source}\nNext: Archive with `python ./.trellis/scripts/task.py archive {task_dir.name}` or start a new task"
has_context = False
for jsonl_name in ("implement.jsonl", "check.jsonl", "spec.jsonl"):
jsonl_path = task_dir / jsonl_name
if jsonl_path.is_file() and _has_curated_jsonl_entry(jsonl_path):
has_context = True
break
has_prd = (task_dir / "prd.md").is_file()
if not has_prd:
return f"Status: NOT READY\nTask: {task_title}\nSource: {active.source}\nMissing: prd.md not created\nNext: Write PRD (see workflow.md Phase 1.1) then curate implement.jsonl per Phase 1.3"
if not has_context:
return f"Status: NOT READY\nTask: {task_title}\nSource: {active.source}\nMissing: implement.jsonl / check.jsonl missing or empty\nNext: Curate entries per workflow.md Phase 1.3 (spec + research files only), then `task.py start`"
return (
f"Status: READY\nTask: {task_title}\n"
f"Source: {active.source}\n"
"Next required action: dispatch `trellis-implement` per Phase 2.1. "
"For agent-capable platforms, the default is to NOT edit code in the main session. "
"After implementation, dispatch `trellis-check` per Phase 2.2 before reporting completion.\n"
"Sub-agent self-exemption: if you are reading this as a `trellis-implement` or "
"`trellis-check` sub-agent (your own role / agent name reflects that), this dispatch "
"instruction does NOT apply to you — you are already the dispatched sub-agent. "
"Implement / check directly without spawning another sub-agent of the same kind.\n"
"User override (per-turn escape hatch): if the user's CURRENT message explicitly tells the "
"main session to handle it directly (\"你直接改\" / \"别派 sub-agent\" / \"main session 写就行\" / "
"\"do it inline\" / \"不用 sub-agent\"), honor it for this turn and edit code directly. "
"Per-turn only; do NOT invent an override the user did not say."
)
def _extract_range(content: str, start_header: str, end_header: str) -> str:
"""Extract lines starting at `## start_header` up to (but excluding) `## end_header`."""
lines = content.splitlines()
start: "int | None" = None
end: int = len(lines)
start_match = f"## {start_header}"
end_match = f"## {end_header}"
for i, line in enumerate(lines):
stripped = line.strip()
if start is None and stripped == start_match:
start = i
continue
if start is not None and stripped == end_match:
end = i
break
if start is None:
return ""
return "\n".join(lines[start:end]).rstrip()
_BREADCRUMB_TAG_RE = re.compile(
r"\[workflow-state:([A-Za-z0-9_-]+)\]\s*\n.*?\n\s*\[/workflow-state:\1\]",
re.DOTALL,
)
def _strip_breadcrumb_tag_blocks(content: str) -> str:
return _BREADCRUMB_TAG_RE.sub("", content)
def _build_workflow_toc(workflow_path: Path) -> str:
"""Inject workflow guide: TOC + Phase Index + Phase 1/2/3 step details.
Since v0.5.0-rc.0 the [workflow-state:STATUS] breadcrumb tag blocks
live inside ## Phase Index. They're consumed by inject-workflow-state.py
on each UserPromptSubmit, so strip them from the session-start payload
to avoid duplicating context.
"""
content = read_file(workflow_path)
if not content:
return "No workflow.md found"
out_lines = [
"# Development Workflow — Section Index",
"Full guide: .trellis/workflow.md (read on demand)",
"",
"## Table of Contents",
]
for line in content.splitlines():
if line.startswith("## "):
out_lines.append(line)
out_lines += ["", "---", ""]
phases = _extract_range(content, "Phase Index", "Customizing Trellis (for forks)")
if phases:
out_lines.append(_strip_breadcrumb_tag_blocks(phases).rstrip())
return "\n".join(out_lines).rstrip()
def main() -> None:
if should_skip_injection():
sys.exit(0)
# Read hook input from stdin
try:
hook_input = json.loads(sys.stdin.read())
if not isinstance(hook_input, dict):
hook_input = {}
project_dir = Path(_normalize_windows_shell_path(hook_input.get("cwd", "."))).resolve()
except (json.JSONDecodeError, KeyError):
hook_input = {}
project_dir = Path(".").resolve()
configure_project_encoding(project_dir)
trellis_dir = project_dir / ".trellis"
context_key = _resolve_context_key(project_dir, hook_input)
output = StringIO()
output.write(SUB_AGENT_NOTICE)
output.write("\n\n")
output.write("""<session-context>
You are starting a new session in a Trellis-managed project.
Read and follow all instructions below carefully.
</session-context>
""")
output.write(FIRST_REPLY_NOTICE)
output.write("\n\n")
output.write("<current-state>\n")
context_script = trellis_dir / "scripts" / "get_context.py"
output.write(run_script(context_script, context_key))
output.write("\n</current-state>\n\n")
output.write("<workflow>\n")
output.write(_build_workflow_toc(trellis_dir / "workflow.md"))
output.write("\n</workflow>\n\n")
output.write("<guidelines>\n")
output.write(
"Project spec indexes are listed by path below. Each index contains a "
"**Pre-Development Checklist** listing the specific guideline files to "
"read before coding.\n\n"
"- If you're spawning an implement/check sub-agent, context is injected "
"automatically via `{task}/implement.jsonl` / `check.jsonl`. You do NOT "
"need to read these indexes yourself.\n"
"- For agent-capable platforms, the default is to dispatch "
"`trellis-implement` and `trellis-check` (so JSONL context is loaded by "
"the sub-agents) rather than editing code in the main session. "
"Honor a per-turn user override only if the user's current message "
"explicitly opts out (see <task-status> below for override phrases).\n"
"- Sub-agent self-exemption: if you are reading this as a `trellis-implement` "
"or `trellis-check` sub-agent, the \"dispatch trellis-implement / trellis-check\" "
"rule above does NOT apply to you — you are already the dispatched sub-agent. "
"Do NOT spawn another sub-agent of the same kind; implement / check directly.\n\n"
)
# guides/ inlined (cross-package thinking, broadly useful)
guides_index = trellis_dir / "spec" / "guides" / "index.md"
if guides_index.is_file():
output.write("## guides (inlined — cross-package thinking guides)\n")
output.write(read_file(guides_index))
output.write("\n\n")
# Other indexes — paths only
paths: list[str] = []
spec_dir = trellis_dir / "spec"
if spec_dir.is_dir():
for sub in sorted(spec_dir.iterdir()):
if not sub.is_dir() or sub.name.startswith("."):
continue
if sub.name == "guides":
continue
index_file = sub / "index.md"
if index_file.is_file():
paths.append(f".trellis/spec/{sub.name}/index.md")
else:
for nested in sorted(sub.iterdir()):
if not nested.is_dir():
continue
nested_index = nested / "index.md"
if nested_index.is_file():
paths.append(
f".trellis/spec/{sub.name}/{nested.name}/index.md"
)
if paths:
output.write("## Available spec indexes (read on demand)\n")
for p in paths:
output.write(f"- {p}\n")
output.write("\n")
output.write(
"Discover more via: "
"`python ./.trellis/scripts/get_context.py --mode packages`\n"
)
output.write("</guidelines>\n\n")
task_status = _get_task_status(trellis_dir, hook_input)
output.write(f"<task-status>\n{task_status}\n</task-status>\n\n")
output.write("""<ready>
Context loaded. Workflow index, project state, and guidelines are already injected above — do NOT re-read them.
When the user sends the first message, follow <task-status> and the workflow guide.
If a task is READY, execute its Next required action without asking whether to continue.
</ready>""")
context = output.getvalue()
result = {
"suppressOutput": True,
"systemMessage": f"Trellis context injected ({len(context)} chars)",
"hookSpecificOutput": {
"hookEventName": "SessionStart",
"additionalContext": context,
},
}
print(json.dumps(result, ensure_ascii=False), flush=True)
if __name__ == "__main__":
main()