@mindfoldhq/trellis 0.6.0-beta.2 → 0.6.0-beta.20

package/dist/templates/markdown/spec/guides/cross-platform-thinking-guide.md.txt

@@ -26,37 +26,68 @@
 | `python3` command | ✅ Always available | ⚠️ May need `python` |
 | `python` command | ⚠️ May be Python 2 | ✅ Usually Python 3 |
 
-**Rule 1**: Always use explicit `python3` in documentation, help text, and error messages.
+**Rule 1**: For user-facing docs, help text, and error messages, either:
+
+- state the platform rule explicitly (`python` on Windows, `python3` elsewhere), or
+- render the command through the same platform-aware helper / placeholder the code uses.
 
 ```python
 # BAD - Assumes shebang works
 print("Usage: ./script.py <args>")
 print("Run: script.py <args>")
 
-# GOOD - Explicit interpreter
-print("Usage: python3 script.py <args>")
-print("Run: python3 ./script.py <args>")
+# GOOD - Platform-aware wording
+print("Usage: python on Windows, python3 elsewhere")
+print("Run: {{PYTHON_CMD}} ./.trellis/scripts/task.py <args>")
 ```
 
-**Rule 2**: When calling Python from TypeScript/Node.js, detect the available command:
+**Rule 2**: When generating config files at init time, use placeholder + platform detection:
 
 ```typescript
+// In template file (settings.json):
+{ "command": "{{PYTHON_CMD}} .claude/hooks/script.py" }
+
+// In configurator:
 function getPythonCommand(): string {
+  return process.platform === "win32" ? "python" : "python3";
+}
+
+function replacePlaceholders(content: string): string {
+  return content.replace(/\{\{PYTHON_CMD\}\}/g, getPythonCommand());
+}
+```
+
+**Rule 3**: When calling Python at runtime from JavaScript, detect platform dynamically:
+
+```javascript
+import { platform } from "os"
+
+const PYTHON_CMD = platform() === "win32" ? "python" : "python3"
+execSync(`${PYTHON_CMD} "${scriptPath}"`, { ... })
+```
+
+**Rule 4**: If you need to verify Python is actually installed (not just choose
+the command), probe the same platform-selected alias you will later render or
+execute:
+
+```typescript
+function getPythonCommand(platform = process.platform): string {
+  return platform === "win32" ? "python" : "python3";
+}
+
+function warnIfPythonTooOld(): void {
+  const cmd = getPythonCommand();
   try {
-    execSync("python3 --version", { stdio: "pipe" });
-    return "python3";
+    execSync(`${cmd} --version`, { stdio: "pipe" });
   } catch {
-    try {
-      execSync("python --version", { stdio: "pipe" });
-      return "python";
-    } catch {
-      return "python3"; // Default, will fail with clear error
-    }
+    // Missing Python is a separate error path; don't silently swap aliases.
   }
 }
 ```
 
-**Rule 3**: When calling Python from Python, use `sys.executable`:
+**Rule 5**: Don't assume the Python version the AI CLI uses matches your shell's `python3`. The user's terminal may resolve `python3` → homebrew 3.11, but AI CLI hosts (including enterprise-forked Claude Code / Cursor distributions) spawn hook subprocesses with a minimal PATH that resolves `python3` → `/usr/bin/python3` → macOS system 3.9. Distributed templates must either target the lowest plausible version or use `from __future__ import annotations` for PEP 604 syntax. See `cli/backend/script-conventions.md` → **CRITICAL: PEP 604 Annotations Require `from __future__ import annotations`** for the hard rule and audit check.
+
+**Rule 6**: When calling Python from Python, use `sys.executable`:
 
 ```python
 import sys
@@ -69,30 +100,6 @@ subprocess.run(["python3", "other_script.py"])
 subprocess.run([sys.executable, "other_script.py"])
 ```
 
-**Rule 4**: Don't assume the Python version your AI CLI uses matches your shell's `python3`. Your terminal may resolve `python3` → 3.11 (via homebrew/pyenv), but AI CLI hosts often spawn hook subprocesses with a minimal PATH that resolves `python3` → the system Python (3.9 on macOS). Any `.py` file run as an AI-CLI hook must be written for the lowest plausible Python version.
-
-Concrete failure: PEP 604 union syntax (`str | None`) requires Python 3.10+. If your hook file uses it, start with `from __future__ import annotations` so annotations become lazy strings and work on Python 3.7+:
-
-```python
-#!/usr/bin/env python3
-"""My hook."""
-from __future__ import annotations     # REQUIRED for PEP 604 annotations
-
-def handler(x: str | None) -> dict | None:   # OK — lazy annotation
-    ...
-```
-
-```python
-# BAD — crashes on Python < 3.10:
-#   TypeError: unsupported operand type(s) for |: 'type' and 'NoneType'
-def handler(x: str | None) -> dict | None:
-    ...
-```
-
-Note: `from __future__ import annotations` only covers **annotations**. Runtime expressions like `isinstance(x, int | str)` still require Python 3.10+. Avoid them in hook scripts.
-
-Applies to anything the AI CLI executes as a hook: `match/case` statements (3.10+), `tomllib` (3.11+), `ExceptionGroup` / `except*` (3.11+) — all crash on older Python regardless of `__future__`.
-
 ### 2. Path Handling
 
 | Assumption | macOS/Linux | Windows |
@@ -101,7 +108,7 @@ Applies to anything the AI CLI executes as a hook: `match/case` statements (3.10
 | `\` separator | ❌ Escape char | ✅ Native |
 | `pathlib.Path` | ✅ Works | ✅ Works |
 
-**Rule**: Use `pathlib.Path` for all path operations.
+**Rule (Python)**: Use `pathlib.Path` for all path operations.
 
 ```python
 # BAD - String concatenation
@@ -112,6 +119,51 @@ from pathlib import Path
 path = Path(base) / filename
 ```
 
+#### Logical key vs filesystem path (TypeScript)
+
+A path string has two distinct roles. **Treat them differently.**
+
+| Role | OS-native (`\` on Windows) | Always POSIX (`/`) |
+|------|---------------------------|--------------------|
+| `fs.readFileSync(p)` / `path.join(cwd, x)` for fs call | ✅ Required | ❌ May fail on Windows |
+| `Map<relPath, content>` key, JSON field, hash dictionary key, anything persisted across OS | ❌ Cross-OS mismatch | ✅ Required |
+
+**Rule**: Anywhere a path string crosses OS or persists (Map keys consumed by another OS, JSON fields, hash dictionary keys), normalize to POSIX. Anywhere it goes straight to `fs.*`, leave OS-native.
+
+**Single source of truth**: `packages/cli/src/utils/posix.ts` exports `toPosix(p)`. Don't sprinkle `replaceAll('\\', '/')` at every `path.join` site — apply `toPosix` **once at the boundary**: collector exit (Map key entering hash dictionary) or write-time (`saveHashes` before `JSON.stringify`).
+
+```typescript
+// BAD - logical key carries OS-native separator
+function collectTemplates(): Map<string, string> {
+  const files = new Map<string, string>();
+  for (const entry of walk(dir)) {
+    files.set(path.join(".opencode", entry), readFile(entry));  // \ on Windows
+  }
+  return files;
+}
+
+// GOOD - normalize at the boundary
+import { toPosix } from "../utils/posix.js";
+
+function collectTemplates(): Map<string, string> {
+  const files = new Map<string, string>();
+  for (const entry of walk(dir)) {
+    files.set(toPosix(path.join(".opencode", entry)), readFile(entry));
+  }
+  return files;
+}
+
+// ALSO ACCEPTABLE - write-side defense (for storage helpers like saveHashes)
+function saveHashes(cwd: string, hashes: Record<string, string>): void {
+  const normalized = Object.fromEntries(
+    Object.entries(hashes).map(([k, v]) => [toPosix(k), v])
+  );
+  fs.writeFileSync(getHashesPath(cwd), JSON.stringify(normalized, null, 2));
+}
+```
+
+**Common offender**: `path.relative(cwd, fullPath)` produces `\` on Windows. If you then use that string as a hash dictionary lookup key (`hashes[relPath]`), `toPosix` it first, or the lookup misses on Windows.
+
 ### 3. Line Endings
 
 | Format | macOS/Linux | Windows | Git |
@@ -119,7 +171,7 @@ path = Path(base) / filename
 | `\n` (LF) | ✅ Native | ⚠️ Some tools | ✅ Normalized |
 | `\r\n` (CRLF) | ⚠️ Extra char | ✅ Native | Converted |
 
-**Rule**: Use `.gitattributes` to enforce consistent line endings.
+**Rule 1**: Use `.gitattributes` to enforce consistent line endings.
 
 ```gitattributes
 * text=auto eol=lf
@@ -127,6 +179,23 @@ path = Path(base) / filename
 *.py text eol=lf
 ```
 
+**Rule 2**: When hashing or comparing **content** across platforms, normalize line endings before computing the hash. `.gitattributes` only governs git checkout — files written by users, scripts, or `core.autocrlf=true` may still arrive as CRLF, and `sha256(LF)` ≠ `sha256(CRLF)` for otherwise-identical content.
+
+```typescript
+// BAD - Windows users with autocrlf=true get a different hash
+export function computeHash(content: string): string {
+  return createHash("sha256").update(content, "utf-8").digest("hex");
+}
+
+// GOOD - normalize before hashing so logical content hashes identically
+export function computeHash(content: string): string {
+  const normalized = content.replace(/\r\n/g, "\n");
+  return createHash("sha256").update(normalized, "utf-8").digest("hex");
+}
+```
+
+Apply this rule wherever the hash crosses OS boundaries (template hash dictionary, content fingerprints stored in JSON, integrity checks against a remote registry).
+
 ### 4. Environment Variables
 
 | Variable | macOS/Linux | Windows |
@@ -135,7 +204,7 @@ path = Path(base) / filename
 | `PATH` separator | `:` | `;` |
 | Case sensitivity | ✅ Case-sensitive | ❌ Case-insensitive |
 
-**Rule**: Use `pathlib.Path.home()` instead of environment variables.
+**Rule 1**: Use `pathlib.Path.home()` instead of environment variables.
 
 ```python
 # BAD
@@ -145,6 +214,31 @@ home = os.environ.get("HOME")
 home = Path.home()
 ```
 
+**Rule 2**: When injecting environment variables into shell commands, generate
+the prefix for the actual shell that will parse the command. Do not choose
+syntax from OS alone. AI tool "Bash" surfaces on Windows may execute through
+PowerShell, Git Bash, MSYS2, or another POSIX-like shell.
+
+```javascript
+// BAD - breaks when the host shell is PowerShell
+command = `export TRELLIS_CONTEXT_ID=${shellQuote(contextKey)}; ${command}`;
+
+// GOOD - shell-dialect-aware command prefix
+const prefix = process.platform === "win32" && !isWindowsPosixShell(process.env)
+  ? `$env:TRELLIS_CONTEXT_ID = ${powershellQuote(contextKey)}; `
+  : `export TRELLIS_CONTEXT_ID=${shellQuote(contextKey)}; `;
+command = `${prefix}${command}`;
+```
+
+On Windows, treat `MSYSTEM`, `MINGW_PREFIX`, `OSTYPE=msys|mingw|cygwin`,
+`SHELL=...bash`, or a platform-specific Git Bash setting as POSIX-shell
+signals. Keep PowerShell as the Windows default when there is no POSIX-shell
+signal.
+
+Also make duplicate-injection detection shell-aware. A guard that only matches
+`export VAR=` will miss PowerShell's `$env:VAR = ...` form and can wrap an
+already-correct command a second time.
+
 ### 5. Command Availability
 
 | Command | macOS/Linux | Windows |
@@ -173,6 +267,25 @@ def tail_follow(file_path: Path) -> None:
                 time.sleep(0.1)
 ```
 
+### Optional Advisory Checks in Agent Sandboxes
+
+AI CLI subprocesses may run with outbound network disabled even when the user's
+normal terminal has network access. Prefer local CLI probes over optional
+network probes when the local CLI already exposes the needed information.
+
+**Rule 1**: Do not let a failed optional advisory check consume a once-per-session
+marker. Write the marker only after the script resolves a usable value and can
+make the intended decision. Otherwise a transient sandbox/network failure hides
+the hint for the rest of the session.
+
+**Rule 2**: If a local command can provide the needed value, try it with a short
+timeout and captured output. For example, `trellis --version` already runs the
+CLI's version comparison logic and can support an actionable update prompt
+without duplicating npm registry parsing.
+
+**Rule 3**: Keep advisory checks silent on failure. The user-facing context output
+must not fail or become noisy because an advisory check could not complete.
+
 ### 6. File Encoding
 
 | Default Encoding | macOS/Linux | Windows |
@@ -183,6 +296,9 @@ def tail_follow(file_path: Path) -> None:
 
 **Rule**: Always explicitly specify `encoding="utf-8"` and use `errors="replace"`.
 
+> **Checklist**: When writing scripts that print non-ASCII, did you configure stdout encoding?
+> See `backend/script-conventions.md` for the specific pattern.
+
 ```python
 # BAD - Relies on system default
 with open(file, "r") as f:
@@ -223,6 +339,12 @@ result = subprocess.run(
 
 When making platform-related changes, check **all these locations**:
 
+### Commands / Skills Sync
+- [ ] New command/skill added to ALL platforms (claude, cursor, iflow, codex, and any new platform)
+- [ ] Each platform's test file updated with new entry in `EXPECTED_COMMAND_NAMES` / `EXPECTED_SKILL_NAMES`
+- [ ] Platform-integration spec's required command table updated if adding a new required command
+- [ ] Command format matches platform convention (see `platform-integration.md` → Command Format by Platform)
+
 ### Documentation & Help Text
 - [ ] Docstrings at top of Python files
 - [ ] `--help` output / argparse descriptions
@@ -239,7 +361,7 @@ When making platform-related changes, check **all these locations**:
 ```bash
 # Find all places that might need updating
 grep -r "python [a-z]" --include="*.py" --include="*.md"
-grep -r "\./" --include="*.py" --include="*.md" | grep -v python3
+grep -r "{{PYTHON_CMD}}\\|python3\\|python " --include="*.py" --include="*.md"
 ```
 
 ---
@@ -248,10 +370,15 @@ grep -r "\./" --include="*.py" --include="*.md" | grep -v python3
 
 Before committing cross-platform code:
 
-- [ ] All Python invocations use `python3` explicitly (docs) or `sys.executable` (code)
+- [ ] User-facing Python invocations are platform-aware (`python` on Windows, `python3` elsewhere) or use `{{PYTHON_CMD}}`
+- [ ] Python subprocesses from Python use `sys.executable`
 - [ ] All paths use `pathlib.Path`
 - [ ] No hardcoded path separators (`/` or `\`)
+- [ ] Path strings used as logical/persisted keys (Map keys, JSON fields, hash dictionary keys) are normalized via `toPosix()`; `fs.*` calls keep OS-native paths
+- [ ] Content hashes computed across OSes normalize line endings (`\r\n` → `\n`) before hashing
+- [ ] Cross-OS JSON with potential legacy pollution carries a `__version` sentinel and the loader discards unknown/legacy versions
 - [ ] No platform-specific commands without fallbacks (e.g., `tail -f`)
+- [ ] Optional advisory checks do not burn once-per-session markers on failure
 - [ ] All file I/O specifies `encoding="utf-8"` and `errors="replace"`
 - [ ] All subprocess calls specify `encoding="utf-8"` and `errors="replace"`
 - [ ] Git commands use `-c i18n.logOutputEncoding=UTF-8`
@@ -283,6 +410,101 @@ output = {
 
 ---
 
+## Cross-Platform Persisted JSON: Schema Migration Sentinel
+
+When a JSON file may be read/written across OSes (committed to git, synced via cloud, copied between machines) **and an older format may already exist on user disks with cross-platform pollution** (Windows-style keys, CRLF-derived hashes, locale-encoded strings), add a `__version` sentinel and let the loader discard old formats so the writer regenerates clean data.
+
+**Why not migrate-in-place?** Path-key migration (`\\` → `/`) plus hash-input migration (CRLF → LF re-hash) plus encoding fixes are correlated — trying to translate the old payload risks producing wrong values. Discarding and regenerating is **safe**: the data is recomputable from disk, and `loadX` returning `{}` triggers the existing init/update path to rebuild canonical entries.
+
+```typescript
+const SCHEMA_VERSION = 2;
+type StoredV2 = { __version: number; hashes: Record<string, string> };
+
+export function loadHashes(cwd: string): Record<string, string> {
+  const file = getHashesPath(cwd);
+  if (!fs.existsSync(file)) return {};
+
+  try {
+    const parsed = JSON.parse(fs.readFileSync(file, "utf-8")) as unknown;
+
+    // Reject legacy flat format (no __version) and unknown versions.
+    // The next saveHashes / initializeHashes will write a fresh v2 file.
+    if (
+      !parsed ||
+      typeof parsed !== "object" ||
+      (parsed as StoredV2).__version !== SCHEMA_VERSION ||
+      typeof (parsed as StoredV2).hashes !== "object"
+    ) {
+      return {};
+    }
+    return (parsed as StoredV2).hashes;
+  } catch {
+    return {};
+  }
+}
+
+export function saveHashes(cwd: string, hashes: Record<string, string>): void {
+  const payload: StoredV2 = { __version: SCHEMA_VERSION, hashes };
+  fs.writeFileSync(getHashesPath(cwd), JSON.stringify(payload, null, 2));
+}
+```
+
+**When to apply**:
+- Hash dictionaries / content fingerprints (e.g., `.template-hashes.json`)
+- Cache files where stale entries are recomputable from authoritative source
+- Any cross-OS persisted file where format change correlates with cross-platform fixes
+
+**When NOT to apply** — if losing the data hurts the user (task state, drafts, settings the user typed). Use real migration there. Sentinel + discard is only safe when data is recomputable.
+
+**Reference**: `packages/cli/src/utils/template-hash.ts` v2 envelope.
+
+---
+
+## JSON/External Data Defensive Checks
+
+When parsing JSON or external data, TypeScript types are **compile-time only**. Runtime data may not match.
+
+**Rule**: Always add defensive checks for required fields before using them.
+
+```typescript
+// BAD - Trusts TypeScript type definition
+interface MigrationItem {
+  from: string;  // TypeScript says required
+  to?: string;
+}
+
+function process(item: MigrationItem) {
+  const path = item.from;  // Runtime: could be undefined!
+}
+
+// GOOD - Defensive check before use
+function process(item: MigrationItem) {
+  if (!item.from) return;  // Skip invalid data
+  const path = item.from;  // Now guaranteed
+}
+```
+
+**When to apply**:
+- Parsing JSON files (manifests, configs)
+- API responses
+- User input
+- Any data from external sources
+
+**Pattern**: Check existence → then use
+
+```typescript
+// Filter pattern - skip invalid items
+const validItems = items.filter(item => item.from && item.to);
+
+// Early return pattern - bail on invalid
+if (!data.requiredField) {
+  console.warn("Missing required field");
+  return defaultValue;
+}
+```
+
+---
+
 ## Common Mistakes
 
 ### 1. "It works on my Mac"
@@ -318,6 +540,9 @@ python3 script.py  # Works!
 # User's Windows (Python from python.org)
 python3 script.py  # 'python3' is not recognized
 python script.py   # Works!
+
+# Trellis docs/config should say the rule, not guess one alias everywhere
+{{PYTHON_CMD}} script.py
 ```
 
 ### 5. "UTF-8 is the default everywhere"
@@ -328,6 +553,9 @@ subprocess.run(cmd, capture_output=True, text=True)  # Works!
 
 # User's Windows (GBK/CP1252 default)
 subprocess.run(cmd, capture_output=True, text=True)  # Garbled Chinese/Unicode
+```
+
+> **Note**: stdout encoding is also affected. See `backend/script-conventions.md` for the fix.
 
 ---
 
@@ -341,3 +569,65 @@ subprocess.run(cmd, capture_output=True, text=True)  # Garbled Chinese/Unicode
 ---
 
 **Core Principle**: If it's not explicit, it's an assumption. And assumptions break.
+
+---
+
+## Release Checklist: Versioned Files
+
+When releasing a new version, ensure **all versioned files** are created/updated:
+
+- [ ] `src/migrations/manifests/{version}.json` - Migration manifest exists
+- [ ] Manifest has correct version, description, changelog
+- [ ] `pnpm build` copies manifests to `dist/`
+- [ ] Test upgrade path from older versions (not just adjacent)
+
+**Why this matters**: Missing manifests cause "path undefined" errors when users upgrade from older versions.
+
+```bash
+# Verify all expected manifests exist
+ls src/migrations/manifests/
+
+# Test upgrade path
+node -e "
+const { getMigrationsForVersion } = require('./dist/migrations/index.js');
+console.log('From 0.2.12:', getMigrationsForVersion('0.2.12', 'CURRENT').length);
+"
+```
+
+## Release Checklist: Bundled Assets
+
+When release notes or docs claim an asset is bundled, installed automatically, or
+included with Trellis, verify the whole distribution path:
+
+- [ ] Source file exists in the branch being tagged, not only in another branch,
+  docs submodule, or marketplace tree.
+- [ ] `pnpm build` copies the asset into `dist/templates/**`.
+- [ ] `npm pack --dry-run --json` includes the expected `dist/**` path.
+- [ ] The built binary installs the asset in a fresh temp repository.
+- [ ] `.trellis/.template-hashes.json` tracks the generated asset path.
+- [ ] `trellis update --dry-run` reports `Already up to date!` in that temp
+  repository.
+
+**Why this matters**: docs/changelog text can move independently from the code
+branch that owns distributable templates. A feature can be documented as bundled
+while the published npm tarball still lacks the files.
+
+```bash
+pnpm --filter @mindfoldhq/trellis build
+
+cd packages/cli
+npm pack --dry-run --json | grep 'dist/templates/common/bundled-skills/<skill>/SKILL.md'
+cd ../..
+
+tmpdir=$(mktemp -d /tmp/trellis-built-bin-smoke-XXXXXX)
+printf '{"name":"trellis-smoke","version":"0.0.0"}\n' > "$tmpdir/package.json"
+git -C "$tmpdir" init -q
+(
+  cd "$tmpdir"
+  node /path/to/Trellis/packages/cli/bin/trellis.js init -u smoke --yes --claude --codex
+  test -f .claude/skills/<skill>/SKILL.md
+  test -f .agents/skills/<skill>/SKILL.md
+  grep -q '<skill>' .trellis/.template-hashes.json
+  node /path/to/Trellis/packages/cli/bin/trellis.js update --dry-run
+)
+```

package/dist/templates/markdown/spec/guides/index.md.txt

@@ -34,6 +34,8 @@ These guides help you **ask the right questions before coding**.
 - [ ] Data format changes between layers
 - [ ] Multiple consumers need the same data
 - [ ] You're not sure where to put some logic
+- [ ] You are adding an event kind, JSONL record, RPC payload, or config field
+- [ ] UI / command code starts casting raw payload fields directly
 
 → Read [Cross-Layer Thinking Guide](./cross-layer-thinking-guide.md)
 
@@ -44,9 +46,25 @@ These guides help you **ask the right questions before coding**.
 - [ ] You're adding a new field to multiple places
 - [ ] **You're modifying any constant or config**
 - [ ] **You're creating a new utility/helper function** ← Search first!
+- [ ] Two files read the same untyped payload field with local casts
+- [ ] Multiple branches update the same derived state from `kind` / `action`
 
 → Read [Code Reuse Thinking Guide](./code-reuse-thinking-guide.md)
 
+### When Verifying AI Cross-Review Results
+
+- [ ] Reviewer claims "user input can be malicious" → Check the actual data source (internal manifest? user config? external API?)
+- [ ] Reviewer flags "missing validation" → Is the data from a trusted internal source?
+- [ ] Reviewer says "behavior change" → Read the code comments — is it intentional design?
+- [ ] Reviewer identifies a "bug" in test → Mentally delete the feature being tested — does the test still pass? If yes → tautological test
+
+**Common AI reviewer false-positive patterns**:
+1. **Trust boundary confusion**: Treating internal data (bundled JSON manifests) as untrusted external input
+2. **Ignoring design comments**: Flagging intentional behavior documented in code comments as bugs
+3. **Variable misreading**: Not tracing a variable to its actual definition (e.g., Map keyed by path vs name)
+
+**Verification rule**: Every CRITICAL/WARNING finding must be verified against the actual code before prioritizing. Budget ~35% false-positive rate for AI reviews.
+
 ---
 
 ## Pre-Modification Rule (CRITICAL)

package/dist/templates/opencode/agents/trellis-check.md

@@ -27,21 +27,25 @@ You are already the `trellis-check` sub-agent that the main session dispatched.
 
 Look for the `<!-- trellis-hook-injected -->` marker in your input above.
 
-- **If the marker is present**: prd / spec / research files have already been auto-loaded for you above. Proceed with the check work directly.
-- **If the marker is absent**: hook injection didn't fire (Windows + Claude Code, `--continue` resume, fork distribution, hooks disabled, etc.). Find the active task path from your dispatch prompt's first line `Active task: <path>` (or run `python3 ./.trellis/scripts/task.py current --source` as a fallback), then Read `<task-path>/prd.md` and the spec files listed in `<task-path>/check.jsonl` yourself before doing the work.
+- **If the marker is present**: task artifacts, spec, and research files have already been auto-loaded for you above. Proceed with the check work directly.
+- **If the marker is absent**: hook injection didn't fire (Windows + Claude Code, `--continue` resume, fork distribution, hooks disabled, etc.). Find the active task path from your dispatch prompt's first line `Active task: <path>` (or run `python3 ./.trellis/scripts/task.py current --source` as a fallback), then Read `<task-path>/check.jsonl`, each listed file, `<task-path>/prd.md`, `<task-path>/design.md` if present, and `<task-path>/implement.md` if present before doing the work.
 
 ## Context
 
 Before checking, read:
 - `.trellis/spec/` - Development guidelines
+- Task `prd.md` - Requirements document
+- Task `design.md` - Technical design (if exists)
+- Task `implement.md` - Execution plan (if exists)
 - Pre-commit checklist for quality standards
 
 ## Core Responsibilities
 
 1. **Get code changes** - Use git diff to get uncommitted code
-2. **Check against specs** - Verify code follows guidelines
-3. **Self-fix** - Fix issues yourself, not just report them
-4. **Run verification** - typecheck and lint
+2. **Review task artifacts** - Check changes against prd.md, design.md if present, and implement.md if present
+3. **Check against specs** - Verify code follows guidelines
+4. **Self-fix** - Fix issues yourself, not just report them
+5. **Run verification** - typecheck and lint
 
 ## Important
 
@@ -60,10 +64,12 @@ git diff --name-only  # List changed files
 git diff              # View specific changes
 ```
 
-### Step 2: Check Against Specs
+### Step 2: Check Against Specs and Task Artifacts
 
-Read relevant specs in `.trellis/spec/` to check code:
+Read the task's prd.md, design.md if present, and implement.md if present, then read relevant specs in `.trellis/spec/` to check code:
 
+- Does it satisfy the task requirements
+- Does it follow the technical design and implementation plan when present
 - Does it follow directory structure conventions
 - Does it follow naming conventions
 - Does it follow code patterns

package/dist/templates/opencode/agents/trellis-implement.md

@@ -27,8 +27,8 @@ You are already the `trellis-implement` sub-agent that the main session dispatch
 
 Look for the `<!-- trellis-hook-injected -->` marker in your input above.
 
-- **If the marker is present**: prd / spec / research files have already been auto-loaded for you above. Proceed with the implementation work directly.
-- **If the marker is absent**: hook injection didn't fire (Windows + Claude Code, `--continue` resume, fork distribution, hooks disabled, etc.). Find the active task path from your dispatch prompt's first line `Active task: <path>` (or run `python3 ./.trellis/scripts/task.py current --source` as a fallback), then Read `<task-path>/prd.md`, `<task-path>/info.md` (if it exists), and the spec files listed in `<task-path>/implement.jsonl` yourself before doing the work.
+- **If the marker is present**: task artifacts, spec, and research files have already been auto-loaded for you above. Proceed with the implementation work directly.
+- **If the marker is absent**: hook injection didn't fire (Windows + Claude Code, `--continue` resume, fork distribution, hooks disabled, etc.). Find the active task path from your dispatch prompt's first line `Active task: <path>` (or run `python3 ./.trellis/scripts/task.py current --source` as a fallback), then Read `<task-path>/implement.jsonl`, each listed file, `<task-path>/prd.md`, `<task-path>/design.md` if present, and `<task-path>/implement.md` if present before doing the work.
 
 ## Context
 
@@ -36,13 +36,14 @@ Before implementing, read:
 - `.trellis/workflow.md` - Project workflow
 - `.trellis/spec/` - Development guidelines
 - Task `prd.md` - Requirements document
-- Task `info.md` - Technical design (if exists)
+- Task `design.md` - Technical design (if exists)
+- Task `implement.md` - Execution plan (if exists)
 
 ## Core Responsibilities
 
 1. **Understand specs** - Read relevant spec files in `.trellis/spec/`
-2. **Understand requirements** - Read prd.md and info.md
-3. **Implement features** - Write code following specs and design
+2. **Understand task artifacts** - Read prd.md, design.md if present, and implement.md if present
+3. **Implement features** - Write code following specs and task artifacts
 4. **Self-check** - Ensure code quality
 5. **Report results** - Report completion status
 
@@ -68,15 +69,15 @@ Read relevant specs based on task type:
 
 ### 2. Understand Requirements
 
-Read the task's prd.md and info.md:
+Read the task's prd.md, design.md if present, and implement.md if present:
 
 - What are the core requirements
 - Key points of technical design
-- Which files to modify/create
+- Implementation order, validation commands, and rollback points
 
 ### 3. Implement Features
 
-- Write code following specs and technical design
+- Write code following specs and task artifacts
 - Follow existing code patterns
 - Only do what's required, no over-engineering