#!/usr/bin/env python3 # -*- coding: utf-8 -*- """ Session Start Hook - Inject structured context """ from __future__ import annotations # IMPORTANT: Suppress all warnings FIRST import warnings warnings.filterwarnings("ignore") import json import os import re import shlex import subprocess import sys from io import StringIO from pathlib import Path 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) return f"{drive}:\\{rest.replace('/', '\\')}" # 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) return f"{drive}:\\{rest.replace('/', '\\')}" # 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) return f"{drive}:\\{rest.replace('/', '\\')}" return path_str 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. """ # IMPORTANT: Force stdout to use UTF-8 on Windows # This fixes UnicodeEncodeError when outputting non-ASCII characters if sys.platform.startswith("win"): import io as _io if hasattr(sys.stdout, "reconfigure"): sys.stdout.reconfigure(encoding="utf-8", errors="replace") # type: ignore[union-attr] elif hasattr(sys.stdout, "detach"): sys.stdout = _io.TextIOWrapper(sys.stdout.detach(), encoding="utf-8", errors="replace") # type: ignore[union-attr] 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 hook-inject and pull-based sub-agent context loaders. """ 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 should_skip_injection() -> bool: """Check if any platform's non-interactive flag is set, or if Trellis hooks are explicitly disabled via TRELLIS_HOOKS=0 / TRELLIS_DISABLE_HOOKS=1. """ if os.environ.get("TRELLIS_HOOKS") == "0": return True if os.environ.get("TRELLIS_DISABLE_HOOKS") == "1": return True non_interactive_vars = [ "CLAUDE_NON_INTERACTIVE", "QODER_NON_INTERACTIVE", "CODEBUDDY_NON_INTERACTIVE", "FACTORY_NON_INTERACTIVE", "CURSOR_NON_INTERACTIVE", "GEMINI_NON_INTERACTIVE", "KIRO_NON_INTERACTIVE", "COPILOT_NON_INTERACTIVE", ] return any(os.environ.get(var) == "1" for var in non_interactive_vars) def read_file(path: Path, fallback: str = "") -> str: try: return path.read_text(encoding="utf-8") except (FileNotFoundError, PermissionError): return fallback 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_context_key(trellis_dir: Path, input_data: dict) -> str | None: 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_context_key # type: ignore[import-not-found] return resolve_context_key(input_data, platform=_detect_platform(input_data)) def _persist_context_key_for_bash(context_key: str | None) -> None: """Expose Trellis session identity to later Claude Code Bash commands. Claude Code SessionStart hooks can append exports to CLAUDE_ENV_FILE; those variables are then available to Bash tools in the same conversation. Without this bridge, `task.py start` has hook stdin during SessionStart but no session identity when the AI later runs it as a normal shell command. """ if not context_key: return env_file = os.environ.get("CLAUDE_ENV_FILE") if not env_file: return try: with open(env_file, "a", encoding="utf-8") as handle: handle.write(f"export TRELLIS_CONTEXT_ID={shlex.quote(context_key)}\n") except OSError: pass def _resolve_active_task(trellis_dir: Path, input_data: 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, input_data, platform=_detect_platform(input_data), ) def run_script(script_path: Path, context_key: str | None = None) -> str: try: if script_path.suffix == ".py": # Add PYTHONIOENCODING to force UTF-8 in subprocess 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)] else: env = os.environ.copy() if context_key: env["TRELLIS_CONTEXT_ID"] = context_key cmd = [str(script_path)] result = subprocess.run( cmd, capture_output=True, text=True, encoding="utf-8", errors="replace", timeout=5, cwd=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, input_data: dict) -> str: """Check current task status and return structured status string with explicit next action. Returns a block with three fields: - Status: current state - Task: task identifier (when applicable) - Next-Action: explicit skill/command/tool call the AI should invoke """ active = _resolve_active_task(trellis_dir, input_data) # Case 1: No active task — waiting for user to describe intent if not active.task_path: return ( "Status: NO ACTIVE TASK\n" f"Source: {active.source}\n" "Next-Action: After the user describes their intent, load skill `trellis-brainstorm` " "to clarify requirements and create a task via `python3 ./.trellis/scripts/task.py create`.\n" "Research reminder: for research-heavy tasks (comparing tools, reading external docs, " "cross-platform surveys), spawn `trellis-research` sub-agents via the Task tool — " "they persist findings to `{TASK_DIR}/research/*.md` and keep main context clean. " "Do NOT do 10+ inline WebFetch/WebSearch in the main conversation.\n" "User override (per-turn escape hatch): if the user's first message explicitly opts " "out of the workflow (\"跳过 trellis\" / \"别走流程\" / \"小修一下\" / \"直接改\" / " "\"skip trellis\" / \"no task\" / \"just do it\"), honor it for this turn — " "acknowledge briefly and proceed without creating a task. Per-turn only." ) # Case 2: Stale pointer — task dir was deleted 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}\n" f"Source: {active.source}\n" f"Next-Action: Run `python3 ./.trellis/scripts/task.py finish` to clear the stale pointer, " "then ask the user what to work on next." ) # Read task.json task_json_path = task_dir / "task.json" task_data = {} 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") # Case 3: Task completed — time to archive if task_status == "completed": return ( f"Status: COMPLETED\nTask: {task_title}\n" f"Source: {active.source}\n" f"Next-Action: Load skill `trellis-update-spec` to capture learnings, " f"then archive with `python3 ./.trellis/scripts/task.py archive {task_dir.name}`." ) has_prd = (task_dir / "prd.md").is_file() # Case 4: No PRD — still in Plan phase if not has_prd: return ( f"Status: PLANNING\nTask: {task_title}\n" f"Source: {active.source}\n" "Next-Action: Load skill `trellis-brainstorm` to clarify requirements with the user " "and produce prd.md in the task directory.\n" "Research reminder: when the task needs external research (tool comparison, docs, " "conventions survey), spawn `trellis-research` sub-agents — don't WebFetch/WebSearch " "inline in the main session. Findings go to `{task_dir}/research/*.md`; PRD only links to them." ) # Case 4b: PRD exists but implement.jsonl has only seed (no curated entries) — Phase 1.3 gate implement_jsonl = task_dir / "implement.jsonl" if implement_jsonl.is_file() and not _has_curated_jsonl_entry(implement_jsonl): return ( f"Status: PLANNING (Phase 1.3)\nTask: {task_title}\n" f"Source: {active.source}\n" "Next-Action: Curate `implement.jsonl` and `check.jsonl` with the spec + research files " "the Phase 2 sub-agents will need. Only spec paths (`.trellis/spec/**/*.md`) and research " "files (`{TASK_DIR}/research/*.md`) — no code paths. Run " "`python3 ./.trellis/scripts/get_context.py --mode packages` to list available specs, " "then edit the jsonl files or use `python3 ./.trellis/scripts/task.py add-context`. " "See `.trellis/workflow.md` Phase 1.3 for details." ) # Case 5: PRD + curated jsonl (or agent-less platform with no jsonl) — enter Execute phase 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 roster: `trellis-implement` (writes code), `trellis-check` (verifies + self-fixes), " "`trellis-research` (persists findings to `research/*.md` — use when you'd otherwise do " "multiple WebFetch/WebSearch inline).\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 _load_trellis_config(trellis_dir: Path, input_data: dict) -> tuple: """Load Trellis config for session-start decisions. Returns: (is_mono, packages_dict, spec_scope, task_pkg, default_pkg) """ scripts_dir = trellis_dir / "scripts" if str(scripts_dir) not in sys.path: sys.path.insert(0, str(scripts_dir)) try: from common.config import get_default_package, get_packages, get_spec_scope, is_monorepo # type: ignore[import-not-found] from common.paths import get_current_task # type: ignore[import-not-found] repo_root = trellis_dir.parent is_mono = is_monorepo(repo_root) packages = get_packages(repo_root) or {} scope = get_spec_scope(repo_root) # Get active task's package task_pkg = None current = get_current_task( repo_root, input_data, platform=_detect_platform(input_data), ) if current: task_json = repo_root / current / "task.json" if task_json.is_file(): try: data = json.loads(task_json.read_text(encoding="utf-8")) if isinstance(data, dict): tp = data.get("package") if isinstance(tp, str) and tp: task_pkg = tp except (json.JSONDecodeError, OSError): pass default_pkg = get_default_package(repo_root) return is_mono, packages, scope, task_pkg, default_pkg except Exception: return False, {}, None, None, None def _check_legacy_spec(trellis_dir: Path, is_mono: bool, packages: dict) -> str | None: """Check for legacy spec directory structure in monorepo. Returns warning message if legacy structure detected, None otherwise. """ if not is_mono or not packages: return None spec_dir = trellis_dir / "spec" if not spec_dir.is_dir(): return None # Check for legacy flat spec dirs (spec/backend/, spec/frontend/ with index.md) has_legacy = False for legacy_name in ("backend", "frontend"): legacy_dir = spec_dir / legacy_name if legacy_dir.is_dir() and (legacy_dir / "index.md").is_file(): has_legacy = True break if not has_legacy: return None # Check which packages are missing spec// directory missing = [ name for name in sorted(packages.keys()) if not (spec_dir / name).is_dir() ] if not missing: return None # All packages have spec dirs if len(missing) == len(packages): return ( f"[!] Legacy spec structure detected: found `spec/backend/` or `spec/frontend/` " f"but no package-scoped `spec//` directories.\n" f"Monorepo packages: {', '.join(sorted(packages.keys()))}\n" f"Please reorganize: `spec/backend/` -> `spec//backend/`" ) return ( f"[!] Partial spec migration detected: packages {', '.join(missing)} " f"still missing `spec//` directory.\n" f"Please complete migration for all packages." ) def _resolve_spec_scope( is_mono: bool, packages: dict, scope, task_pkg: str | None, default_pkg: str | None, ) -> set | None: """Resolve which packages should have their specs injected. Returns: Set of package names to include, or None for full scan. """ if not is_mono or not packages: return None # Single-repo: full scan if scope is None: return None # No scope configured: full scan if isinstance(scope, str) and scope == "active_task": if task_pkg and task_pkg in packages: return {task_pkg} if default_pkg and default_pkg in packages: return {default_pkg} return None # Fallback to full scan if isinstance(scope, list): valid = set() for entry in scope: if entry in packages: valid.add(entry) else: print( f"Warning: spec_scope contains unknown package: {entry}, ignoring", file=sys.stderr, ) if valid: # Warn if active task is out of scope if task_pkg and task_pkg not in valid: print( f"Warning: active task package '{task_pkg}' is out of configured spec_scope", file=sys.stderr, ) return valid # All entries invalid: fallback chain print( "Warning: all spec_scope entries invalid, falling back to task/default/full", file=sys.stderr, ) if task_pkg and task_pkg in packages: return {task_pkg} if default_pkg and default_pkg in packages: return {default_pkg} return None # Full scan return None # Unknown scope type: full scan def _extract_range(content: str, start_header: str, end_header: str) -> str: """Extract lines starting at `## start_header` up to (but excluding) `## end_header`. Both parameters are full header lines WITHOUT the `## ` prefix (e.g. "Phase Index"). Returns empty string if start header is not found. End header missing → extracts to end of file. """ 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: """Remove `[workflow-state:STATUS]...[/workflow-state:STATUS]` blocks. The tag blocks live inside `## Phase Index` (since v0.5.0-rc.0, when they were colocated with their phase summaries) and are consumed by the UserPromptSubmit hook (`inject-workflow-state.py`). The session-start payload already covers the full step bodies, so re-inlining the breadcrumbs here would just duplicate context. """ return _BREADCRUMB_TAG_RE.sub("", content) def _build_workflow_overview(workflow_path: Path) -> str: """Inject the workflow guide for the session. Contents: 1. Section index (all `## ` headings — navigation) 2. Phase Index section (rules, skill routing table, anti-rationalization table) 3. Phase 1/2/3 step-level details (the actual how-to for each step) The meta sections (Core Principles / Trellis System / Customizing Trellis) are NOT injected — Core Principles is short prose the AI can Read on demand; Trellis System lists reference commands duplicated in step bodies; Customizing Trellis is for forks. Workflow-state breadcrumb tag blocks (which now live inside Phase Index since v0.5.0-rc.0) are stripped from the extracted range — they're consumed by the UserPromptSubmit hook, not the session-start preamble. Total budget: Phase Index ~2 KB + Phase 1/2/3 ~7 KB = ~9 KB. """ 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 += ["", "---", ""] # Extract Phase Index through the end of Phase 3 (before "Customizing # Trellis" — the docs-for-forks footer added in v0.5.0-rc.0). Since # sections appear in order Phase Index → Phase 1 → Phase 2 → Phase 3 → # Customizing Trellis, a single range grab captures all four. The # breadcrumb tag blocks now embedded inside Phase Index are stripped so # they don't duplicate the per-turn UserPromptSubmit injection. 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(): if should_skip_injection(): sys.exit(0) try: hook_input = json.loads(sys.stdin.read()) if not isinstance(hook_input, dict): hook_input = {} except (json.JSONDecodeError, ValueError): hook_input = {} # Try platform-specific env vars, hook cwd, fallback to cwd project_dir_env_vars = [ "CLAUDE_PROJECT_DIR", "QODER_PROJECT_DIR", "CODEBUDDY_PROJECT_DIR", "FACTORY_PROJECT_DIR", "CURSOR_PROJECT_DIR", "GEMINI_PROJECT_DIR", "KIRO_PROJECT_DIR", "COPILOT_PROJECT_DIR", ] project_dir = None for var in project_dir_env_vars: val = os.environ.get(var) if val: project_dir = Path(_normalize_windows_shell_path(val)).resolve() break if project_dir is None: project_dir = Path(_normalize_windows_shell_path(hook_input.get("cwd", "."))).resolve() trellis_dir = project_dir / ".trellis" context_key = _resolve_context_key(trellis_dir, hook_input) _persist_context_key_for_bash(context_key) # Load config for scope filtering and legacy detection is_mono, packages, scope_config, task_pkg, default_pkg = _load_trellis_config( trellis_dir, hook_input, ) allowed_pkgs = _resolve_spec_scope(is_mono, packages, scope_config, task_pkg, default_pkg) output = StringIO() output.write(""" You are starting a new session in a Trellis-managed project. Read and follow all instructions below carefully. """) output.write(FIRST_REPLY_NOTICE) output.write("\n\n") # Legacy migration warning legacy_warning = _check_legacy_spec(trellis_dir, is_mono, packages) if legacy_warning: output.write(f"\n{legacy_warning}\n\n\n") output.write("\n") context_script = trellis_dir / "scripts" / "get_context.py" output.write(run_script(context_script, context_key)) output.write("\n\n\n") output.write("\n") output.write(_build_workflow_overview(trellis_dir / "workflow.md")) output.write("\n\n\n") output.write("\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 " "or loaded by the sub-agent 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 below for override phrases).\n\n" ) # guides/ is cross-package thinking — always include inline (small, 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 spec indexes — paths only (main agent reads on demand; # sub-agents get their specific specs via jsonl injection) 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 # already inlined above index_file = sub / "index.md" if index_file.is_file(): # Flat spec dir (single-repo layer like spec/backend/) paths.append(f".trellis/spec/{sub.name}/index.md") else: # Nested package dirs (monorepo: spec///index.md) # Apply scope filter if allowed_pkgs is not None and sub.name not in allowed_pkgs: continue 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: " "`python3 ./.trellis/scripts/get_context.py --mode packages`\n" ) output.write("\n\n") # Check task status and inject structured tag task_status = _get_task_status(trellis_dir, hook_input) output.write(f"\n{task_status}\n\n\n") output.write(""" 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 and the workflow guide. If a task is READY, execute its Next required action without asking whether to continue. """) result = { "hookSpecificOutput": { "hookEventName": "SessionStart", "additionalContext": output.getvalue(), } } # Output JSON - stdout is already configured for UTF-8 print(json.dumps(result, ensure_ascii=False), flush=True) if __name__ == "__main__": main()