@blxzer/cursor-trellis 0.2.9 → 0.3.2

package/dist/templates/common/skills/brainstorm.md

@@ -1,165 +1,189 @@
-# Trellis Brainstorm
-
-## Non-Negotiable Interview Contract
-
-Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one-by-one. For each question, provide your recommended answer.
-
-Ask the questions one at a time.
-
-## Non-Negotiable Evidence Rule
-
-If a question can be answered by exploring the codebase, explore the codebase instead.
-
-This is mandatory. Before asking the user a question, first check whether the answer is already available in code, tests, configs, docs, existing specs, or task history.
-
-Do not ask the user to confirm facts that the repository can answer. Ask only for product intent, preference, scope, risk tolerance, or decisions that remain ambiguous after inspection.
-
----
-
-Use this skill during Phase 1 planning to turn the user's request into clear requirements and planning artifacts.
-
-**Agent-capable platforms:** Do **not** use legacy Claude-only grill subagents as a hard gate. Complete **PRD Grill** (below) and **`trellis-micro-grill`** for blocking open questions before treating planning as ready for `design.md` / `implement.md` / `start-execution --check`.
-
-## Preconditions
-
-Use this skill only after task-creation consent has been given and the user is ready to enter Trellis planning.
-
-If no task exists yet, create one:
-
-```bash
-TASK_DIR=$(python ./.trellis/scripts/task.py create "<short task title>" --slug <slug>)
-```
-
-Use a concise title from the user's request. Use a slug without a date prefix. `task.py create` adds the `MM-DD-` directory prefix automatically.
-
-`task.py create` creates the default `prd.md`. Update that file with the current understanding before asking follow-up questions.
-
-## Two-phase planning overview
-
-| Phase | Name | User questions |
-| --- | --- | --- |
-| **A** | Discovery Before Questions + PRD draft | None until repo evidence is exhausted |
-| **B** | PRD Grill pass + Micro-grill unresolved | Only blocking business / risk / preference |
-
-External facts during Discovery or Research: load `smart-search-cli`; on CLI/doctor failure use Cursor WebSearch/WebFetch and persist under `{TASK}/research/` with `source: cursor-web-fallback`.
-
----
-
-## Phase A — Discovery Before Questions
-
-Run **before** any user interview questions.
-
-Inspect and record in `prd.md` (sections: **Confirmed facts**, initial **Out of scope**, draft **Goal**):
-
-1. **Code & tests** — relevant modules, fixtures, configs, error paths.
-2. **Specs** — `.trellis/spec/` indexes and layer guides for touched packages.
-3. **History** — archived tasks, active task research, developer journal when useful.
-4. **Platform** — project platform hooks, agents, and skills (see `.trellis/spec/guides/cursor-subagent-policy.md` when using Cursor); shared `.agents/skills/` when behavior spans platforms.
-5. **Parent/Child** — if multiple independent deliverables, note child split early in `prd.md`.
-
-Use retrieval per `.trellis/spec/guides/retrieval-daily-guide.md` (rg for literals, codegraph for structure, fast-context for semantic sweep).
-
-Dispatch **`trellis-research`** (writable Agent) when a topic needs a dedicated `{TASK}/research/<topic>.md` file; do **not** use a subagent for PRD Grill itself.
-
-## Phase A — PRD draft
-
-After Discovery, flesh out `prd.md`:
-
-- goal and user value
-- confirmed facts (not restated as unverified requirements)
-- requirements
-- draft acceptance criteria
-- out of scope
-- open questions (tag **blocking** vs **nice-to-have**)
-
-For complex tasks, start `design.md` / `implement.md` skeletons only when boundaries are already clear from Discovery; otherwise wait until Phase B.
-
-## Phase B — PRD Grill pass
-
-Treat `prd.md` (+ existing `design.md` fragments) as the **only document surface**. Run this checklist; fix the PRD in place (no new subagent):
-
-| # | Check |
-| --- | --- |
-| 1 | **Goal & user value** — single clear statement |
-| 2 | **Confirmed facts vs assumptions** — repo facts not listed as assumptions |
-| 3 | **Testable acceptance criteria** |
-| 4 | **Out of scope** explicit |
-| 5 | **Dependencies & sequencing** |
-| 6 | **Parent/Child & deliverables** when applicable |
-| 7 | **Research & external facts** — smart-search or documented fallback |
-| 8 | **Execution gate & artifacts** — `design.md` / `implement.md` / `verify.md` expectations |
-| 9 | **Durable Learning** — Phase 3.3 will need `update-spec` \| `no-update` \| `unsure` |
-| 10 | **Platform** — Cursor-first; PRD Grill in-session (no legacy grill-me / grill-with-docs subagent gate) |
-| 11 | **Risk & rollback** for complex tasks |
-| 12 | **Open questions** — only **blocking** strategic/preference items remain |
-
-## Phase B — Micro-grill unresolved
-
-For each **blocking** open question after the checklist, embed the **`trellis-micro-grill` contract**:
-
-- exactly **one** question per message
-- **Simplified Chinese** for user-facing text
-- recommended answer + trade-off
-- **update `prd.md` after every answer** before the next question
-
-Stop micro-grill when no blocking open questions remain.
-
-Do not ask process questions ("should I search?"). Do not re-ask facts Discovery already confirmed.
-
-## Question Rules (Phase B only)
-
-Each question must include:
-
-- the decision needed
-- why the answer matters
-- your recommended answer
-- the trade-off if the user chooses differently
-
-## Artifact Rules
-
-`prd.md` records requirements and acceptance:
-
-- goal and user value
-- confirmed facts
-- requirements
-- acceptance criteria
-- out of scope
-- open questions that still block planning
-
-`design.md` records technical design for complex tasks:
-
-- architecture and boundaries
-- data flow and contracts
-- compatibility and migration notes
-- important trade-offs
-- operational or rollback considerations
-
-`implement.md` records execution planning for complex tasks:
-
-- ordered implementation checklist
-- validation commands
-- risky files or rollback points
-- **Development Strategy Contract** (`execution_mode`, `isolation`, …): before finalizing, run `python3 ./.trellis/scripts/task.py suggest-execution-strategy <task-dir>` (or `--json`); reconcile suggestion with scope, then paste the approved YAML block into `implement.md`
-- follow-up checks before `task.py start-execution --check`
-
-Lightweight tasks may have only `prd.md`. Complex tasks must have `prd.md`, `design.md`, and `implement.md` before `task.py start-execution --check`.
-
-`implement.md` is not a replacement for `implement.jsonl`. Use JSONL files only for manifest-style spec and research references when the task needs them.
-
-## Completion criteria — PRD Grill done
-
-Planning is ready for execution gate when **all** hold:
-
-- PRD Grill checklist (12 items) satisfied or explicitly N/A with rationale in `prd.md`
-- **No blocking** open questions in `prd.md`
-- Acceptance criteria are testable; out of scope is explicit
-- Complex tasks: `design.md` and `implement.md` present
-- User reviewed artifacts or explicitly approved proceeding
-
-Then proceed to Phase 1.2 Research (if needed), Phase 1.4 `task.py start-execution --check`, and implementation only after user approval.
-
-Do not start implementation until the user approves or asks for implementation.
-
-## Legacy planning flow (summary)
-
+# Trellis Brainstorm
+
+## Non-Negotiable Interview Contract
+
+Interview me relentlessly about every aspect of this plan until we reach a shared understanding. Walk down each branch of the design tree, resolving dependencies between decisions one-by-one. For each question, provide your recommended answer.
+
+Ask the questions one at a time.
+
+## Non-Negotiable Evidence Rule
+
+If a question can be answered by exploring the codebase, explore the codebase instead.
+
+This is mandatory. Before asking the user a question, first check whether the answer is already available in code, tests, configs, docs, existing specs, or task history.
+
+Do not ask the user to confirm facts that the repository can answer. Ask only for product intent, preference, scope, risk tolerance, or decisions that remain ambiguous after inspection.
+
+## Thinking Principles
+
+These two principles shape *how* you grill, not just *what* you ask. Apply them throughout Phase A and B.
+
+### First Principles
+
+Reason from irreducible facts and user value, not from analogy to existing solutions or "how it's always been done".
+
+- When you encounter a requirement phrased as "we need X like Y has", separate the **root need** (the user value behind X) from the **inherited assumption** (Y's specific shape).
+- For every "because that's the convention" justification, ask: what problem did that convention solve, and does that problem actually exist here?
+- Trace each requirement back to a user-facing outcome. If no outcome survives, the requirement is inherited, not fundamental — flag it as a candidate for Out of Scope.
+
+### Occam's Razor
+
+Among competing designs that satisfy the acceptance criteria, prefer the one with the fewest additional assumptions.
+
+- Default recommendation = the **minimal sufficient** design. Complexity must be justified by evidence (a contract, an edge case, a verified constraint), not added speculatively "for later".
+- Actively strip scope that exists only to satisfy an assumption you cannot verify from repo evidence.
+- When two paths are equally sufficient, the simpler one wins; the burden of proof is on the more complex path.
+
+---
+
+Use this skill during Phase 1 planning to turn the user's request into clear requirements and planning artifacts.
+
+**Agent-capable platforms:** Do **not** use legacy Claude-only grill subagents as a hard gate. Complete **PRD Grill** (below) and **`cstl-micro-grill`** for blocking open questions before treating planning as ready for `design.md` / `implement.md` / `start-execution --check`.
+
+## Preconditions
+
+Use this skill only after task-creation consent has been given and the user is ready to enter Trellis planning.
+
+If no task exists yet, create one:
+
+```bash
+TASK_DIR=$(python ./.cstl/scripts/task.py create "<short task title>" --slug <slug>)
+```
+
+Use a concise title from the user's request. Use a slug without a date prefix. `task.py create` adds the `MM-DD-` directory prefix automatically.
+
+`task.py create` creates the default `prd.md`. Update that file with the current understanding before asking follow-up questions.
+
+## Two-phase planning overview
+
+| Phase | Name | User questions |
+| --- | --- | --- |
+| **A** | Discovery Before Questions + PRD draft | None until repo evidence is exhausted |
+| **B** | PRD Grill pass + Micro-grill unresolved | Only blocking business / risk / preference |
+
+External facts during Discovery or Research: load `smart-search-cli`; on CLI/doctor failure use Cursor WebSearch/WebFetch and persist under `{TASK}/research/` with `source: cursor-web-fallback`.
+
+---
+
+## Phase A — Discovery Before Questions
+
+Run **before** any user interview questions.
+
+Inspect and record in `prd.md` (sections: **Confirmed facts**, initial **Out of scope**, draft **Goal**):
+
+1. **Code & tests** — relevant modules, fixtures, configs, error paths.
+2. **Specs** — `.cstl/spec/` indexes and layer guides for touched packages.
+3. **History** — archived tasks, active task research, developer journal when useful.
+4. **Platform** — project platform hooks, agents, and skills (see `.cstl/spec/guides/cursor-subagent-policy.md` when using Cursor); shared `.agents/skills/` when behavior spans platforms.
+5. **Parent/Child** — if multiple independent deliverables, note child split early in `prd.md`.
+
+Use retrieval per `.cstl/spec/guides/retrieval-daily-guide.md` (rg for literals, codegraph for structure, fast-context for semantic sweep).
+
+Dispatch **`cstl-research`** (writable Agent) when a topic needs a dedicated `{TASK}/research/<topic>.md` file; do **not** use a subagent for PRD Grill itself.
+
+## Phase A — PRD draft
+
+After Discovery, flesh out `prd.md`:
+
+- goal and user value
+- confirmed facts (not restated as unverified requirements)
+- requirements
+- draft acceptance criteria
+- out of scope
+- open questions (tag **blocking** vs **nice-to-have**)
+
+For complex tasks, start `design.md` / `implement.md` skeletons only when boundaries are already clear from Discovery; otherwise wait until Phase B.
+
+## Phase B — PRD Grill pass
+
+Treat `prd.md` (+ existing `design.md` fragments) as the **only document surface**. Run this checklist; fix the PRD in place (no new subagent):
+
+| # | Check |
+| --- | --- |
+| 1 | **Goal & user value** — single clear statement |
+| 2 | **Confirmed facts vs assumptions** — repo facts not listed as assumptions |
+| 3 | **Testable acceptance criteria** |
+| 4 | **Out of scope** explicit |
+| 5 | **Dependencies & sequencing** |
+| 6 | **Parent/Child & deliverables** when applicable |
+| 7 | **Research & external facts** — smart-search or documented fallback |
+| 8 | **Execution gate & artifacts** — `design.md` / `implement.md` / `verify.md` expectations |
+| 9 | **Durable Learning** — Phase 3.3 will need `update-spec` \| `no-update` \| `unsure` |
+| 10 | **Platform** — Cursor-first; PRD Grill in-session (no legacy grill-me / grill-with-docs subagent gate) |
+| 11 | **Risk & rollback** for complex tasks |
+| 12 | **Open questions** — only **blocking** strategic/preference items remain |
+| 13 | **Root vs inherited assumptions** — every requirement traces to a user-facing root need; inherited assumptions (convention, analogy, "we've always done it") are flagged or removed (First Principles) |
+| 14 | **Minimal sufficient design** — no scope, layer, or mechanism exists to satisfy an unverified assumption; the simplest design that meets acceptance wins (Occam's Razor) |
+
+## Phase B — Micro-grill unresolved
+
+For each **blocking** open question after the checklist, embed the **`cstl-micro-grill` contract**:
+
+- exactly **one** question per message
+- **Simplified Chinese** for user-facing text
+- recommended answer + trade-off
+- **update `prd.md` after every answer** before the next question
+
+Stop micro-grill when no blocking open questions remain.
+
+Do not ask process questions ("should I search?"). Do not re-ask facts Discovery already confirmed.
+
+## Question Rules (Phase B only)
+
+Each question must include:
+
+- the decision needed
+- why the answer matters
+- your recommended answer
+- the trade-off if the user chooses differently
+
+Your recommended answer defaults to Occam's Razor: the **minimal sufficient** option that still satisfies the acceptance criteria. Only recommend a more complex option when you can cite evidence (a contract, a verified edge case, a repo constraint) that the simpler option violates.
+
+## Artifact Rules
+
+`prd.md` records requirements and acceptance:
+
+- goal and user value
+- confirmed facts
+- requirements
+- acceptance criteria
+- out of scope
+- open questions that still block planning
+
+`design.md` records technical design for complex tasks:
+
+- architecture and boundaries
+- data flow and contracts
+- compatibility and migration notes
+- important trade-offs
+- operational or rollback considerations
+
+`implement.md` records execution planning for complex tasks:
+
+- ordered implementation checklist
+- validation commands
+- risky files or rollback points
+- **Development Strategy Contract** (`execution_mode`, `isolation`, …): before finalizing, run `python3 ./.cstl/scripts/task.py suggest-execution-strategy <task-dir>` (or `--json`); reconcile suggestion with scope, then paste the approved YAML block into `implement.md`
+- follow-up checks before `task.py start-execution --check`
+
+Lightweight tasks may have only `prd.md`. Complex tasks must have `prd.md`, `design.md`, and `implement.md` before `task.py start-execution --check`.
+
+`implement.md` is not a replacement for `implement.jsonl`. Use JSONL files only for manifest-style spec and research references when the task needs them.
+
+## Completion criteria — PRD Grill done
+
+Planning is ready for execution gate when **all** hold:
+
+- PRD Grill checklist (14 items) satisfied or explicitly N/A with rationale in `prd.md`
+- **No blocking** open questions in `prd.md`
+- Acceptance criteria are testable; out of scope is explicit
+- Complex tasks: `design.md` and `implement.md` present
+- User reviewed artifacts or explicitly approved proceeding
+
+Then proceed to Phase 1.2 Research (if needed), Phase 1.4 `task.py start-execution --check`, and implementation only after user approval.
+
+Do not start implementation until the user approves or asks for implementation.
+
+## Legacy planning flow (summary)
+
 The former single "Planning Flow" is now Phase A + B above. Steps 4–6 map to Phase B micro-grill and artifact updates.

package/dist/templates/common/skills/break-loop.md

@@ -55,8 +55,8 @@ What broader problems does this bug reveal?
 
 Solidify insights into the system:
 
-- [ ] Update `.trellis/spec/guides/` thinking guides
-- [ ] Update relevant `.trellis/spec/` docs
+- [ ] Update `.cstl/spec/guides/` thinking guides
+- [ ] Update relevant `.cstl/spec/` docs
 - [ ] Create issue record (if applicable)
 - [ ] Create feature ticket for root fix
 - [ ] Update check guidelines if needed
@@ -118,7 +118,7 @@ Three levels of insight:
    - If it's a code reuse issue → update `code-reuse-thinking-guide.md`
    - If it's domain-specific → update `backend/*.md` or `frontend/*.md`
 
-2. **Sync templates** - After updating `.trellis/spec/`, sync to `src/templates/markdown/spec/`
+2. **Sync templates** - After updating `.cstl/spec/`, sync to `src/templates/markdown/spec/`
 
 3. **Commit the spec updates** - This is the primary output, not just the analysis text
 

package/dist/templates/common/skills/check.md

@@ -20,13 +20,13 @@ Read the selected task artifacts in order:
 - `implement.md` if present
 
 ```bash
-python ./.trellis/scripts/get_context.py --mode packages
+python ./.cstl/scripts/get_context.py --mode packages
 ```
 
 For each changed package/layer, read the spec index and follow its **Quality Check** section:
 
 ```bash
-cat .trellis/spec/<package>/<layer>/index.md
+cat .cstl/spec/<package>/<layer>/index.md
 ```
 
 Read the specific guideline files referenced — the index is a pointer, not the goal.
@@ -56,12 +56,12 @@ Run the project's lint, type-check, and test commands. Fix any failures before p
 - [ ] `verify.md` contains exactly one token: `Learning decision: update-spec` | `no-update` | `unsure`
 - [ ] If `update-spec` or `unsure`: `research/learning-proposal.md` exists (or documented `N/A` with reason) and matches the decision
 - [ ] If `no-update`: includes `no durable learning` (or guide-equivalent) plus brief rationale
-- [ ] If `update-spec`: spec was written only after confirmation; `Spec update evidence:` points at `.trellis/spec/...`
-- [ ] No silent edits to `.trellis/spec/` without confirmation
+- [ ] If `update-spec`: spec was written only after confirmation; `Spec update evidence:` points at `.cstl/spec/...`
+- [ ] No silent edits to `.cstl/spec/` without confirmation
 
 ### Spec Sync
 
-- [ ] Does `.trellis/spec/` need updates? (route through semi-automatic flow: proposal → confirm → `trellis-update-spec`)
+- [ ] Does `.cstl/spec/` need updates? (route through semi-automatic flow: proposal → confirm → `cstl-update-spec`)
 
 > "If I fixed a bug or discovered something non-obvious, should I document it so future me won't hit the same issue?" → If YES, update the relevant spec doc.
 

package/dist/templates/common/skills/update-spec.md

@@ -9,9 +9,9 @@ When you learn something valuable (from debugging, implementing, or discussion),
 ## Semi-automatic flow (required for durable learning)
 
 1. **Detect** — did this task produce reusable code-spec or guide-worthy learning?
-2. **Proposal** — write `{TASK}/research/learning-proposal.md` (target spec paths, draft bullets; **do not** edit `.trellis/spec/` yet).
+2. **Proposal** — write `{TASK}/research/learning-proposal.md` (target spec paths, draft bullets; **do not** edit `.cstl/spec/` yet).
 3. **Confirm** — user or task owner explicitly approves (`Learning decision: update-spec` in `verify.md`).
-4. **Write** — only then edit spec files; add `Spec update evidence: .trellis/spec/...` to `verify.md`.
+4. **Write** — only then edit spec files; add `Spec update evidence: .cstl/spec/...` to `verify.md`.
 
 **Forbidden:** silent spec edits; writing spec when `Learning decision: no-update` or `unsure` without follow-up; auto-updating spec from hooks or check alone.
 
@@ -68,7 +68,7 @@ For triggered tasks, include all sections below:
 ## Spec Structure Overview
 
 ```
-.trellis/spec/
+.cstl/spec/
 ├── <layer>/           # Per-layer coding standards (e.g., backend/, frontend/, api/)
 │   ├── index.md       # Overview and links
 │   └── *.md           # Topic-specific guidelines
@@ -130,7 +130,7 @@ Before editing, read the current code-spec to:
 - Find the right section for your update
 
 ```bash
-cat .trellis/spec/<category>/<file>.md
+cat .cstl/spec/<category>/<file>.md
 ```
 
 ### Step 4: Make the Update

package/dist/templates/cursor/agents/{trellis-check.md → cstl-check.md}

@@ -1,5 +1,5 @@
 ---
-name: trellis-check
+name: cstl-check
 description: Trellis quality check agent. Use this exact agent for Trellis task verification, check.jsonl context injection, and self-fixing code review. Do not use generic/default/generalPurpose agents for Trellis checks.
 tools: Read, Write, Edit, Bash, Glob, Grep, mcp__exa__web_search_exa, mcp__exa__get_code_context_exa
 ---
@@ -7,11 +7,11 @@ tools: Read, Write, Edit, Bash, Glob, Grep, mcp__exa__web_search_exa, mcp__exa__
 ## Entry points
 
 - **Agent session:** Open this agent file manually in a new chat — context from this file + your main-session prompt.
-- **Task dispatch:** Run `python ./.trellis/scripts/generate_dispatch_prompt.py --agent check` → pass stdout as `Task(..., prompt=...)` — context from the Layer 2 prompt.
+- **Task dispatch:** Run `python ./.cstl/scripts/generate_dispatch_prompt.py --agent check` → pass stdout as `Task(..., prompt=...)` — context from the Layer 2 prompt.
 
 ## Context source
 
-- **Layer 2 prompt = PRIMARY (guaranteed)** — always generate via CLI before `Task(trellis-check)`.
+- **Layer 2 prompt = PRIMARY (guaranteed)** — always generate via CLI before `Task(cstl-check)`.
 - **Hook `additional_context` = best-effort only** — Cursor #158452: not guaranteed to reach the model; optimization / fallback only.
 
 # Check Agent
@@ -21,19 +21,19 @@ You are the Check Agent in the Trellis workflow.
 ## Model policy
 
 - **Default:** no `model:` → **inherit** parent session.
-- **Per dispatch:** main session asks user → one-shot `model:` overlay → `Task` → restore (architecture review uses the same flow). See `.trellis/spec/guides/cursor-subagent-policy.md`.
+- **Per dispatch:** main session asks user → one-shot `model:` overlay → `Task` → restore (architecture review uses the same flow). See `.cstl/spec/guides/cursor-subagent-policy.md`.
 
 ## Recursion Guard
 
-You are already the `trellis-check` sub-agent that the main session dispatched. Do the review and fixes directly.
+You are already the `cstl-check` sub-agent that the main session dispatched. Do the review and fixes directly.
 
-- Do NOT spawn another `trellis-check` or `trellis-implement` sub-agent.
-- If SessionStart context, workflow-state breadcrumbs, or workflow.md say to dispatch `trellis-implement` / `trellis-check`, treat that as a main-session instruction that is already satisfied by your current role.
+- Do NOT spawn another `cstl-check` or `cstl-implement` sub-agent.
+- If SessionStart context, workflow-state breadcrumbs, or workflow.md say to dispatch `cstl-implement` / `cstl-check`, treat that as a main-session instruction that is already satisfied by your current role.
 - Only the main session may dispatch Trellis implement/check agents. If more implementation work is needed, report that recommendation instead of spawning.
 
 ## Trellis Context Loading Protocol
 
-Look for the `<!-- trellis-hook-injected -->` marker in your input above.
+Look for the `<!-- cstl-hook-injected -->` marker in your input above.
 
 - **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 selected task path from your dispatch prompt's first line `Selected task: <path>`, 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.
@@ -41,13 +41,13 @@ Look for the `<!-- trellis-hook-injected -->` marker in your input above.
 ## Dispatch contract (Parent / inline)
 
 - Only the **main session or Parent** dispatches this agent; Child workers must not re-spawn Trellis sub-agents.
-- **Inline** (`in_progress-inline`): main session uses the `trellis-check` **skill** instead of spawning this agent unless a dedicated review pass is needed.
-- Align with the `trellis-check` skill: `get_context.py --mode packages`, spec indexes, cross-layer checks; you may fix issues and record gates — do not redefine Parent `task-map` or gate semantics.
+- **Inline** (`in_progress-inline`): main session uses the `cstl-check` **skill** instead of spawning this agent unless a dedicated review pass is needed.
+- Align with the `cstl-check` skill: `get_context.py --mode packages`, spec indexes, cross-layer checks; you may fix issues and record gates — do not redefine Parent `task-map` or gate semantics.
 
 ## Context
 
 Before checking, read:
-- `.trellis/spec/` - Development guidelines
+- `.cstl/spec/` - Development guidelines
 - Task `prd.md` - Requirements document
 - Task `design.md` - Technical design (if exists)
 - Task `implement.md` - Execution plan (if exists)
@@ -65,7 +65,7 @@ Before checking, read:
 
 - Reviewer id: `cursor`.
 - When `implement.md` quality_gates requires a reviewer gate for the current transition, write human-readable evidence in `verify.md` before recording the machine-checkable gate result. Parent/Child integration evidence may also belong in Parent `task-map.md`.
-- Record non-baseline gates with `python ./.trellis/scripts/task.py record-gate <task> --transition <transition> --gate <gate> --result PASS --reviewer cursor --evidence verify.md`.
+- Record non-baseline gates with `python ./.cstl/scripts/task.py record-gate <task> --transition <transition> --gate <gate> --result PASS --reviewer cursor --evidence verify.md`.
 - For FAIL, add `--root-cause implementation-defect|contract-changing-defect|validation-environment-blocker` and `--issue-fingerprint <short-stable-id>`. Route implementation defects back to Execution, contract-changing defects to Planning, and validation blockers to Verification / Review.
 - For SKIPPED, use only explicit user approval: `--skip-approved-by user --skip-reason <reason>`.
 - Never record `baseline-check`; the CLI owns it. Do not pass review bodies, logs, screenshots, or long issue lists through `record-gate` arguments.
@@ -89,7 +89,7 @@ git diff              # View specific changes
 
 ### Step 2: Check Against Specs and Task Artifacts
 
-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:
+Read the task's prd.md, design.md if present, and implement.md if present, then read relevant specs in `.cstl/spec/` to check code:
 
 - Does it satisfy the task requirements
 - Does it follow the technical design and implementation plan when present
@@ -124,7 +124,7 @@ When the pack exists:
 
 ```bash
 # Optional: confirm path from selected task
-python ./.trellis/scripts/task.py selected --source
+python ./.cstl/scripts/task.py selected --source
 ```
 
 Read `{TASK}/research/retrieval-pack-latest.json` and use `contextPack.selected` (top items: `title`, `source`, `reference`, `score`). Add or update `verify.md` with:

package/dist/templates/cursor/agents/{trellis-implement.md → cstl-implement.md}

@@ -1,5 +1,5 @@
 ---
-name: trellis-implement
+name: cstl-implement
 description: Trellis implementation agent. Use this exact agent for Trellis task implementation, implement.jsonl context injection, and hook-injection tests. Do not use generic/default/generalPurpose agents for Trellis implementation. No git commit allowed.
 tools: Read, Write, Edit, Bash, Glob, Grep, mcp__exa__web_search_exa, mcp__exa__get_code_context_exa
 ---
@@ -7,11 +7,11 @@ tools: Read, Write, Edit, Bash, Glob, Grep, mcp__exa__web_search_exa, mcp__exa__
 ## Entry points
 
 - **Agent session:** Open this agent file manually in a new chat — context from this file + your main-session prompt.
-- **Task dispatch:** Run `python ./.trellis/scripts/generate_dispatch_prompt.py --agent implement` → pass stdout as `Task(..., prompt=...)` — context from the Layer 2 prompt.
+- **Task dispatch:** Run `python ./.cstl/scripts/generate_dispatch_prompt.py --agent implement` → pass stdout as `Task(..., prompt=...)` — context from the Layer 2 prompt.
 
 ## Context source
 
-- **Layer 2 prompt = PRIMARY (guaranteed)** — always generate via CLI before `Task(trellis-implement)`.
+- **Layer 2 prompt = PRIMARY (guaranteed)** — always generate via CLI before `Task(cstl-implement)`.
 - **Hook `additional_context` = best-effort only** — Cursor #158452: not guaranteed to reach the model; optimization / fallback only.
 
 # Implement Agent
@@ -21,19 +21,19 @@ You are the Implement Agent in the Trellis workflow.
 ## Model policy
 
 - **Default:** no `model:` → **inherit** parent session.
-- **Per dispatch / Child worker:** main session asks user → one-shot `model:` overlay on this file → `Task` → restore. Child session uses the same overlay pattern on `trellis-implement`. See `.trellis/spec/guides/cursor-subagent-policy.md`.
+- **Per dispatch / Child worker:** main session asks user → one-shot `model:` overlay on this file → `Task` → restore. Child session uses the same overlay pattern on `cstl-implement`. See `.cstl/spec/guides/cursor-subagent-policy.md`.
 
 ## Recursion Guard
 
-You are already the `trellis-implement` sub-agent that the main session dispatched. Do the implementation work directly.
+You are already the `cstl-implement` sub-agent that the main session dispatched. Do the implementation work directly.
 
-- Do NOT spawn another `trellis-implement` or `trellis-check` sub-agent.
-- If SessionStart context, workflow-state breadcrumbs, or workflow.md say to dispatch `trellis-implement` / `trellis-check`, treat that as a main-session instruction that is already satisfied by your current role.
+- Do NOT spawn another `cstl-implement` or `cstl-check` sub-agent.
+- If SessionStart context, workflow-state breadcrumbs, or workflow.md say to dispatch `cstl-implement` / `cstl-check`, treat that as a main-session instruction that is already satisfied by your current role.
 - Only the main session may dispatch Trellis implement/check agents. If more parallel work is needed, report that recommendation instead of spawning.
 
 ## Trellis Context Loading Protocol
 
-Look for the `<!-- trellis-hook-injected -->` marker in your input above.
+Look for the `<!-- cstl-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, `/multitask` parallel dispatch, etc.). Find the selected task path from your dispatch prompt's first line `Selected task: <path>`, 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.
@@ -41,20 +41,20 @@ Look for the `<!-- trellis-hook-injected -->` marker in your input above.
 ## Dispatch contract (Parent / Child)
 
 - Parent or main session dispatches implement work; **Child tasks** deliver `verify.md` + `handoff.md` and must not change shared gate contracts.
-- Do not spawn nested `trellis-implement` / `trellis-check`; recommend a Parent review when check is needed.
+- Do not spawn nested `cstl-implement` / `cstl-check`; recommend a Parent review when check is needed.
 
 ## Context
 
 Before implementing, read:
-- `.trellis/workflow.md` - Project workflow
-- `.trellis/spec/` - Development guidelines
+- `.cstl/workflow.md` - Project workflow
+- `.cstl/spec/` - Development guidelines
 - Task `prd.md` - Requirements document
 - 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/`
+1. **Understand specs** - Read relevant spec files in `.cstl/spec/`
 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
@@ -76,8 +76,8 @@ Before implementing, read:
 
 Read relevant specs based on task type:
 
-- Spec layers: `.trellis/spec/<package>/<layer>/`
-- Shared guides: `.trellis/spec/guides/`
+- Spec layers: `.cstl/spec/<package>/<layer>/`
+- Shared guides: `.cstl/spec/guides/`
 
 ### 2. Understand Requirements