trekoon 0.2.6 → 0.2.7

package/README.md

@@ -2,761 +2,95 @@
 
 AI-first issue tracking for humans and agents.
 
-Trekoon is a Bun-powered CLI focused on execution workflows where AI agents and humans share the same task graph.
+Trekoon is a local-first CLI for planning and execution inside a repository. It
+keeps work in a shared **epic → task → subtask** graph so humans and agents can
+work against the same state, from the terminal, with deterministic machine
+output when automation needs it.
 
-## Installation
+## What it is for
 
-Recommended (global install with Bun):
-
-```bash
-bun add -g trekoon
-```
-
-Then verify:
-
-```bash
-trekoon --help
-trekoon quickstart
-```
-
-Alternative (npm global install):
-
-```bash
-npm i -g trekoon
-```
-
-## What Trekoon is
-
-- Local-first CLI issue tracker
-- Structured hierarchy: **epic → task → subtask**
-- UUID-based references for durable linking across branches/worktrees
-- Dependency-aware planning and execution
-- Output modes:
-  - **Human mode** for terminal users
-  - **JSON mode** for stable machine parsing
-  - **TOON mode** for true TOON-encoded payloads
-
-## What Trekoon aims to accomplish
-
-1. Make issue tracking fast enough for daily terminal use.
-2. Make issue data deterministic and machine-readable for AI automation.
-3. Keep one repo-scoped state store that every worktree can coordinate through safely.
-4. Stay minimal in code size while preserving robustness and clear boundaries.
-
-## Command surface
-
-- `trekoon init`
-- `trekoon help [command]`
-- `trekoon quickstart`
-- `trekoon epic <create|expand|list|show|search|replace|update|delete>`
-- `trekoon session`
-- `trekoon task <create|create-many|list|show|ready|next|done|search|replace|update|delete>`
-- `trekoon subtask <create|create-many|list|search|replace|update|delete>`
-- `trekoon dep <add|add-many|remove|list|reverse>`
-- `trekoon events prune [--dry-run] [--archive] [--retention-days <n>]`
-- `trekoon migrate <status|rollback> [--to-version <n>]`
-- `trekoon sync <status|pull|resolve|conflicts>`
-- `trekoon skills install [--link --editor opencode|claude|pi] [--to <path>] [--allow-outside-repo]`
-- `trekoon skills update`
-- `trekoon wipe --yes`
-
-Global output modes:
-
-- `--json` for structured JSON output
-- `--toon` for true TOON-encoded output (not JSON text)
-- `--compat <mode>` for explicit machine compatibility behavior
-- `--help` for root and command help
-- `--version` for CLI version
-
-Global options can be used before or after the command:
-
-```bash
-trekoon --toon quickstart
-trekoon quickstart --toon
-trekoon --json quickstart
-trekoon quickstart --json
-```
-
-Trekoon options use long form (`--option`) for command/subcommand flags.
-Root help/version aliases `-h` and `-v` are also supported.
-
-Human view options:
-
-- List and show commands default to table output in human mode.
-- Use `--view compact` to restore compact pipe output.
-- `epic list`, `task list`, and `subtask list` support `--view table|compact`.
-- `epic show` and `task show` support `--view table|compact|tree|detail`.
-
-List defaults and filters (`epic list`, `task list`, `subtask list`):
-
-- Default scope: open work only (`in_progress`, `in-progress`, `todo`)
-- Default limit: `10`
-- Status filter: `--status in_progress,todo` (CSV)
-- Custom limit: `--limit <n>`
-- Cursor pagination: `--cursor <n>` (offset-like start index for next page)
-- All rows and statuses: `--all`
-- `--all` is mutually exclusive with `--status`, `--limit`, and `--cursor`
-
-Bulk updates (`epic update`, `task update`, `subtask update`):
-
-- Target all rows: `--all`
-- Target specific rows: `--ids <id1,id2,...>`
-- Bulk updates support only `--append <text>` and/or `--status <status>`
-- In bulk mode, do not pass a positional id
-- `--all` and `--ids` are mutually exclusive
-- `--append` and `--description` are mutually exclusive
-
-Examples:
-
-```bash
-trekoon task update --all --status in_progress
-trekoon task update --ids <task-1>,<task-2> --append "\nFollow-up note"
-trekoon subtask update --all --status done
-trekoon subtask update --ids <subtask-1>,<subtask-2> --append "\nFollow-up note"
-trekoon epic update --ids <epic-1>,<epic-2> --status done
-```
-
-## Quickstart
-
-Trekoon is local-first, but in git repos and worktrees it is **repo-shared**:
-every worktree for the same repository resolves to one shared `.trekoon`
-directory and one shared `.trekoon/trekoon.db`.
-
-- `worktreeRoot` identifies the current checkout.
-- `sharedStorageRoot` identifies the repository root that owns `.trekoon`.
-- `databaseFile` points at the shared SQLite database.
-- `.trekoon` stays gitignored on purpose because the DB is operational state,
-  not source code.
-- Committing `.trekoon/trekoon.db` is the wrong fix for drift because it bakes
-  machine-local state and stale snapshots into Git.
-
-Outside git repos, Trekoon falls back to the invocation cwd.
-
-When machine output is enabled (`--json`/`--toon`) and a command resolves
-storage from a non-canonical cwd, Trekoon emits
-`meta.storageRootDiagnostics` so automation can verify the storage contract.
-
-### 1) Initialize
-
-```bash
-trekoon init
-trekoon --version
-```
-
-Bootstrap expectations:
-
-- Run `trekoon --toon init` once per repository to create or re-bootstrap the
-  shared storage root.
-- Run `trekoon --toon sync status` before agent work to inspect diagnostics.
-- If diagnostics report `recoveryRequired`, a tracked/ignored mismatch, or an
-  ambiguous recovery path, stop and repair setup before continuing.
-- Do **not** continue with task selection after broken bootstrap warnings.
-
-### 2) Create epic → task → subtask
-
-```bash
-trekoon epic create --title "Agent backlog stabilization" --description "Track stabilization work" --status todo
-trekoon task create --title "Implement sync status" --description "Add status reporting" --epic <epic-id> --status todo
-trekoon subtask create --task <task-id> --title "Add cursor model" --status todo
-trekoon task list
-trekoon task list --status done
-trekoon task list --limit 25
-trekoon task list --all --view compact
-```
-
-### 2a) Preferred one-shot epic creation
-
-When you already know the epic tree, create the epic, tasks, subtasks, and
-dependencies in one invocation.
-
-```bash
-trekoon epic create \
-  --title "Batch command rollout" \
-  --description "Ship one-shot planning workflows" \
-  --task "task-a|First task|First description|todo" \
-  --task "task-b|Second task|Second description|todo" \
-  --subtask "@task-a|sub-a|First subtask|Subtask description|todo" \
-  --dep "@task-b|@task-a" \
-  --dep "@sub-a|@task-a"
-```
-
-Use this when:
-
-- the epic does not exist yet
-- later records need to reference earlier created records via `@temp-key`
-- you want one atomic create step and one machine response with mappings/counts
-
-Compact machine output adds:
-
-```text
-command: epic.create
-data:
-  epic: created epic row
-  tasks[]: created tasks in input order
-  subtasks[]: created subtasks in input order
-  dependencies[]: created dependencies in input order
-  result:
-    mappings[]: { kind: task|subtask, tempKey, id }
-    counts: { tasks, subtasks, dependencies }
-```
-
-### 3) Add dependencies
-
-```bash
-trekoon dep add <task-id> <depends-on-id>
-trekoon dep list <task-id>
-```
-
-### 3a) Batch planning commands
-
-Use compact batch commands when one invocation needs to create or link multiple
-items atomically. Use the single-item commands when you already have persisted
-UUIDs and only need one mutation.
-
-#### `task create-many`
-
-Create multiple tasks under one epic in declared order.
-
-```bash
-trekoon task create-many \
-  --epic <epic-id> \
-  --task "seed-api|Design API|Define batch grammar|todo" \
-  --task "seed-cli|Wire CLI|Hook parser and output|in_progress"
-```
-
-Compact spec:
-
-- `--task <temp-key>|<title>|<description>|<status>`
-- escape `\|`, `\\`, `\n`, `\r`, `\t`
-- repeated `--task` flags are preserved in the exact order provided
-- temp keys are local mapping labels, not persisted IDs
-
-Rollback semantics:
-
-- Trekoon validates the full batch before inserts
-- duplicate temp keys, empty required fields, or invalid input fail the whole
-  command
-- no partial task rows are kept on failure
-
-Compact machine output:
-
-```text
-command: task.create-many
-data:
-  epicId: <epic-id>
-  tasks[]: created task rows in input order
-  result:
-    mappings[]: { kind: task, tempKey, id }
-```
-
-#### `subtask create-many`
-
-Create multiple subtasks under one existing task.
-
-```bash
-trekoon subtask create-many <task-id> \
-  --subtask "seed-tests|Write tests|Cover happy path|todo" \
-  --subtask "seed-docs|Document flow|Add operator notes|todo"
-```
-
-Equivalent explicit parent form:
-
-```bash
-trekoon subtask create-many \
-  --task <task-id> \
-  --subtask "seed-tests|Write tests|Cover happy path|todo"
-```
-
-Rules:
-
-- positional `<task-id>` or `--task <task-id>` may be used
-- if both are provided, they must be identical or the command fails
-- repeated `--subtask` flags are applied in declared order
-
-Rollback semantics:
-
-- full batch prevalidation happens before inserts
-- duplicate temp keys, conflicting task ids, or invalid specs abort the whole
-  command
-- no partial subtasks are kept on failure
-
-Compact machine output:
-
-```text
-command: subtask.create-many
-data:
-  taskId: <task-id>
-  subtasks[]: created subtask rows in input order
-  result:
-    mappings[]: { kind: subtask, tempKey, id }
-```
-
-#### `dep add-many`
-
-Create multiple dependency edges in one ordered, transactional operation.
-
-```bash
-trekoon dep add-many \
-  --dep "<task-b>|<task-a>" \
-  --dep "<subtask-c>|<task-b>"
-```
-
-Compact spec:
-
-- `--dep <source-ref>|<depends-on-ref>`
-- repeated `--dep` flags are applied in declared order
-- standalone `dep add-many` resolves persisted IDs only
-- `@temp-key` refs are **not** resolved from earlier commands; they are reserved
-  for same-invocation workflows such as `epic expand`
-
-Rollback semantics:
-
-- validation covers the full dependency set before insert
-- missing ids, unresolved `@temp-key` refs, duplicates, or cycles fail the whole
-  batch
-- no partial dependency edges are inserted on failure
-
-Compact machine output:
+- Fast issue tracking for day-to-day terminal use
+- One repo-scoped task graph that works across branches and worktrees
+- Stable machine-readable output for AI workflows (`--toon`, `--json`)
+- Minimal command surface with strong planning and execution primitives
 
-```text
-command: dep.add-many
-data:
-  dependencies[]: created dependency rows in input order
-  result:
-    mappings[]: []
-```
-
-#### `epic expand`
-
-Expand one existing epic by creating tasks, subtasks, and dependencies in one
-transaction. Use this when the epic already exists and you want to add a linked
-batch later.
-
-```bash
-trekoon epic expand <epic-id> \
-  --task "task-api|Design API|Define compact grammar|todo" \
-  --task "task-cli|Wire CLI|Hook parser and output|todo" \
-  --subtask "@task-api|sub-tests|Write tests|Cover parser cases|todo" \
-  --dep "@task-cli|@task-api" \
-  --dep "@sub-tests|@task-api"
-```
-
-Compact specs:
-
-- `--task <temp-key>|<title>|<description>|<status>`
-- `--subtask <parent-ref>|<temp-key>|<title>|<description>|<status>`
-- `--dep <source-ref>|<depends-on-ref>`
-- `@temp-key` refs may target tasks/subtasks declared earlier in the same
-  `epic expand` invocation
-
-Background phases:
-
-1. validate all compact specs and duplicate temp keys
-2. create tasks transactionally
-3. resolve subtask parent temp keys and create subtasks
-4. resolve dependency refs and link dependencies
-5. append task, subtask, then dependency events
-6. roll back the full expansion if any phase fails
-
-Compact machine output:
-
-```text
-command: epic.expand
-data:
-  epicId: <epic-id>
-  tasks[]: created tasks in input order
-  subtasks[]: created subtasks in input order
-  dependencies[]: created dependencies in input order
-  result:
-    mappings[]: { kind: task|subtask, tempKey, id }
-    counts: { tasks, subtasks, dependencies }
-```
-
-When to choose which command:
-
-- use `task create-many` for sibling tasks under one known epic
-- use `subtask create-many` for sibling subtasks under one known task
-- use `dep add-many` only when every endpoint already has a persisted ID
-- use `epic create` with batch specs when the epic does not exist yet and the
-  whole graph is known up front
-- use `epic expand` when the epic already exists and one batch must add linked
-  tasks/subtasks/dependencies with `@temp-key` references
-
-### 4) AI execution loop for agents
-
-The primary loop is: **session → work → task done → repeat**.
-
-Orient with a single call that returns diagnostics, sync status, next ready
-task with subtasks, blocker list, and readiness counts:
-
-```bash
-trekoon --toon session
-```
-
-Claim work, then finish or report a block:
-
-```bash
-trekoon --toon task update <task-id> --status in_progress
-trekoon --toon task done <task-id>
-trekoon --toon task update <task-id> --append "Blocked by <reason>" --status blocked
-```
-
-`task done` marks the task done and returns the next ready task with
-dependencies inline, replacing the old multi-step transition.
-
-Fail-fast rules:
-
-- Treat `meta.storageRootDiagnostics` as the source of truth for worktree
-  storage.
-- In linked worktrees, `sharedStorageRoot` may differ from `worktreeRoot`; that
-  is expected.
-- If `recoveryRequired` is `true`, stop and follow the reported bootstrap or
-  recovery action.
-- Do not fall back to a separate per-worktree DB or continue after missing
-  shared storage.
-
-<details>
-<summary>Legacy manual bootstrap (use <code>session</code> instead)</summary>
-
-```bash
-trekoon --toon init
-trekoon --toon sync status
-trekoon --toon task ready --limit 5
-trekoon --toon task next
-trekoon --toon dep reverse <task-or-subtask-id>
-trekoon --toon task update <task-id> --status in_progress
-trekoon --toon task update <task-id> --append "Completed implementation and checks" --status done
-```
-
-</details>
-
-### 5) Use TOON output for agent workflows
-
-```bash
-trekoon --toon epic show <epic-id>
-trekoon --toon task show <task-id>
-```
-
-Optional alternative for integrations that explicitly require JSON:
-
-```bash
-trekoon --json epic show <epic-id>
-trekoon --json task show <task-id>
-```
-
-### 6) Scoped search for repeated text
-
-Use scoped search before manual tree reads when you need to locate repeated
-paths, labels, or migration targets.
-
-```bash
-trekoon --toon epic search <epic-id> "path/to/somewhere"
-trekoon --toon task search <task-id> "path/to/somewhere"
-trekoon --toon subtask search <subtask-id> "path/to/somewhere"
-```
-
-Scope rules:
-
-- `epic search` scans the epic title/description plus every task and subtask
-  title/description in that epic tree.
-- `task search` scans the task title/description plus descendant subtask
-  title/description.
-- `subtask search` scans only that subtask's title/description.
-- Add `--fields title`, `--fields description`, or
-  `--fields title,description` when you need a narrower scan.
-
-### 7) Preview first, then apply scoped replace
-
-Use search first to confirm the scope, then run replace in preview mode, and
-only use `--apply` after the preview matches the intended migration.
-
-```bash
-trekoon --toon epic search <epic-id> "path/to/somewhere"
-trekoon --toon epic replace <epic-id> --search "path/to/somewhere" --replace "path/to/new-path"
-trekoon --toon epic replace <epic-id> --search "path/to/somewhere" --replace "path/to/new-path" --apply
-```
-
-Use this loop for low-risk agent workflows:
-
-1. `search` when you need the smallest possible read before deciding whether a
-   migration is needed.
-2. preview `replace` to verify the exact candidate set and changed fields.
-3. `replace --apply` only after the preview output matches the intended scope.
+## Why it exists
 
-Epic-scoped replace applies across the epic title/description and every task and
-subtask title/description in that epic tree.
+Trekoon exists to make task tracking cheap enough to use while coding, and
+structured enough that agents can read, update, and complete work without
+guessing.
 
-Compact TOON expectations for agents:
+## Install
 
-```text
-ok: true
-command: epic.search
-data:
-  scope: epic
-  query: { search, fields[], mode: preview }
-  matches[]: { kind, id, fields[]: { field, count, snippet } }
-  summary: { matchedEntities, matchedFields, totalMatches }
-metadata:
-  contractVersion: 1.0.0
-  requestId: req-<id>
-```
-
-```text
-ok: true
-command: epic.replace
-data:
-  scope: epic
-  query: { search, replace, fields[], mode: preview|apply }
-  matches[]: { kind, id, fields[]: { field, count, snippet } }
-  summary: { matchedEntities, matchedFields, totalMatches, mode }
-metadata:
-  contractVersion: 1.0.0
-  requestId: req-<id>
-```
-
-Background behavior:
-
-- `epic search` and preview `epic replace` traverse the epic first, then
-  descendant tasks, then descendant subtasks.
-- Within each record, Trekoon checks `title` before `description` so output stays
-  deterministic and low-token.
-- Preview reports the candidate set without mutating records.
-- `--apply` reuses the same scoped traversal, updates only rows with real text
-  changes, and returns the matched rows with `query.mode` and `summary.mode`
-  set to `"apply"`.
-
-### 8) Sync workflow for worktrees
-
-- Run `trekoon sync status` at session start and before PR/merge.
-- Run `trekoon sync pull --from main` before merge to align tracker state.
-- If conflicts exist, resolve explicitly:
-
-```bash
-trekoon sync status
-trekoon sync pull --from main
-trekoon sync conflicts list
-trekoon sync conflicts show <conflict-id>
-trekoon sync resolve <conflict-id> --use ours
-```
-
-`sync pull` machine output includes diagnostics counters and hints so agents can
-react deterministically:
-
-- `diagnostics.malformedPayloadEvents`
-- `diagnostics.applyRejectedEvents`
-- `diagnostics.quarantinedEvents`
-- `diagnostics.conflictEvents`
-- `diagnostics.errorHints`
-
-Worktree diagnostics and recovery:
-
-- Inspect `storageMode`, `repoCommonDir`, `worktreeRoot`, `sharedStorageRoot`,
-  and `databaseFile` in machine output when debugging worktree behavior.
-- If a worktree resolves shared storage outside its checkout, that is expected
-  for linked worktrees and should not be “fixed” by committing `.trekoon`.
-- If Git contains a tracked `.trekoon/trekoon.db`, remove it from Git history or
-  the index as appropriate, keep `.trekoon` ignored, and re-run
-  `trekoon --toon init`.
-- Use `trekoon wipe --yes` only for explicit destructive recovery; it deletes
-  the shared storage root for the entire repository, not just the current
-  worktree.
-
-Compatibility mode for legacy sync command IDs:
-
-```bash
-trekoon --json --compat legacy-sync-command-ids sync status
-trekoon --toon --compat legacy-sync-command-ids sync pull --from main
-```
-
-Behavior:
-
-- Default remains strict canonical IDs (`sync.status`, `sync.pull`, ...).
-- Compatibility mode rewrites sync command IDs to legacy forms
-  (`sync_status`, `sync_pull`, ...).
-- Compatibility mode is machine-only and valid only for `sync` commands.
-- Machine output includes `metadata.compatibility` with:
-  - deprecation warning code
-  - migration guidance
-  - canonical + compatibility command IDs
-  - removal window (`removalAfter: 2026-09-30`)
-- Migration path: remove `--compat legacy-sync-command-ids` and consume dotted
-  command IDs directly.
-
-### 9) Install project-local Trekoon skill for agents
-
-`trekoon skills install` always writes the bundled skill file under the current
-working directory at:
-
-- `.agents/skills/trekoon/SKILL.md`
-
-You can also create a project-local editor link:
+Recommended (global install with Bun):
 
 ```bash
-trekoon skills install
-trekoon skills install --link --editor opencode
-trekoon skills install --link --editor claude
-trekoon skills install --link --editor pi
-trekoon skills install --link --editor opencode --to ./.custom-editor/skills
-trekoon skills update
+bun add -g trekoon
 ```
 
-Path behavior:
-
-- Default opencode link path: `.opencode/skills/trekoon`
-- Default claude link path: `.claude/skills/trekoon`
-- Default pi link path: `.pi/skills/trekoon`
-- `--to <path>` overrides the editor root for link creation only.
-- `--to` does **not** move or copy `SKILL.md` to that path.
-- By default, link targets must resolve inside the current working directory root.
-- Editor symlinks are written with relative targets so repo-local links survive
-  moving the repository.
-- Use `--allow-outside-repo` only for intentional external links.
-- When override is used, install prints a warning and includes confirmation
-  fields in machine output.
-- Re-running install is idempotent: it refreshes `SKILL.md` and reuses/replaces
-  the same symlink target.
-- `trekoon skills update` is idempotent: it refreshes canonical
-  `.agents/skills/trekoon/SKILL.md` and creates or refreshes default editor
-  links when their config directories exist.
-- Update skips editors with no config dir and leaves conflicts untouched while
-  reporting actionable path context.
-- If the link destination exists as a non-link path, install fails with an
-  actionable conflict error.
-
-How `--to` works (step-by-step):
-
-1. Trekoon always installs/copies to:
-   - `<cwd>/.agents/skills/trekoon/SKILL.md`
-2. If `--link` is present, Trekoon creates a `trekoon` symlink directory entry.
-3. `--to <path>` sets the symlink root directory.
-4. Final link path is:
-   - `<resolved-to-path>/trekoon -> <relative path to <cwd>/.agents/skills/trekoon>`
-
-Example:
+Alternative (npm global install):
 
 ```bash
-trekoon skills install --link --editor opencode --to ./.custom-editor/skills
-```
-
-This produces:
-
-- `<cwd>/.agents/skills/trekoon/SKILL.md` (copied file)
-- `<cwd>/.custom-editor/skills/trekoon` (symlink)
-- symlink target: relative path to `<cwd>/.agents/skills/trekoon`
-
-Trekoon does not mutate global editor config directories.
-
-### 10) Pre-merge checklist
-
-- [ ] `trekoon sync status` shows no unresolved conflicts
-- [ ] done tasks/subtasks are marked completed
-- [ ] dependency graph has no stale blockers
-- [ ] final AI check: `trekoon --toon epic show <epic-id>`
-- [ ] no one tried to commit `.trekoon/trekoon.db` as a worktree fix
-
-## Machine-contract recipes (--toon)
-
-Use `--toon` for production agent loops. The examples below show command +
-expected envelope fields.
-
-Base envelope fields (all machine responses):
-
-```text
-ok: true|false
-command: <stable command id>
-data: <payload>
-metadata:
-  contractVersion: "1.0.0"
-  requestId: req-<stable-id>
+npm i -g trekoon
 ```
 
-Most subcommand identifiers are dot-namespaced (`task.list`, `sync.pull`,
-`epic.show`). Root-level commands may use single-token IDs (`help`, `init`,
-`quickstart`, `wipe`, `version`).
-
-Additional metadata can appear when relevant:
-
-- `metadata.compatibility` when `--compat` mode is active
-- `meta.storageRootDiagnostics` when a machine-readable command resolves
-  storage from a non-canonical cwd
-
-### Ready queue (deterministic candidates)
+Verify the install:
 
 ```bash
-trekoon --toon task ready --limit 3
+trekoon --help
+trekoon quickstart
 ```
 
-Payload fields:
+## Commands
 
-```text
-ok: true
-command: task.ready
-data:
-  candidates[]:
-    task: { id, epicId, title, status, ... }
-    readiness: { isReady, reason }
-    blockerSummary: { blockedByCount, totalDependencies, blockedBy[] }
-    ranking: { rank, blockerCount, statusPriority }
-  blocked[]: (same shape, non-ready items)
-  summary: {
-    totalOpenTasks,
-    readyCount,
-    returnedCount,
-    appliedLimit,
-    blockedCount,
-    unresolvedDependencyCount,
-  }
-```
+These are the commands most people need to recognize quickly:
 
-### Reverse dependency walk (blocker impact)
+| Goal | Commands |
+| --- | --- |
+| Initialize a repo | `trekoon init` |
+| Learn the CLI | `trekoon help [command]`, `trekoon quickstart` |
+| Plan work | `trekoon epic ...`, `trekoon task ...`, `trekoon subtask ...`, `trekoon dep ...` |
+| Start an execution session | `trekoon session` |
+| Keep worktrees in sync | `trekoon sync ...` |
+| Install or refresh the AI skill | `trekoon skills install`, `trekoon skills update` |
+| Maintenance | `trekoon events prune ...`, `trekoon migrate ...`, `trekoon wipe --yes` |
 
-```bash
-trekoon --toon dep reverse <task-or-subtask-id>
-```
+Machine output modes:
 
-Payload fields:
+- `--toon` for true TOON-encoded payloads
+- `--json` for JSON output
+- `--compat <mode>` for explicit compatibility behavior
+- `--help` and `--version` at the root or command level
 
-```text
-ok: true
-command: dep.reverse
-data:
-  targetId: <id>
-  targetKind: task|subtask
-  blockedNodes[]: { id, kind, distance, isDirect }
-```
+For the full command surface, flags, filters, and bulk update rules, read the
+[command reference](docs/commands.md).
 
-### Pagination contract for machine list calls
+## AI skill
 
-```bash
-trekoon --toon task list --status todo --limit 2
-trekoon --toon task list --status todo --limit 2 --cursor 2
-```
+Trekoon ships with a bundled `trekoon` skill for AI agents. It teaches the
+agent to:
 
-Cursor semantics:
+- use `--toon` by default
+- prefer the smallest sufficient read
+- use bulk planning commands when possible
+- keep progress in Trekoon with append-based updates
+- treat `.trekoon` as shared repo-scoped operational state
 
-- `--cursor <n>` is offset-like pagination for list endpoints (`epic list`,
-  `task list`, `subtask list`).
-- Do not combine `--all` with `--cursor`.
-- Machine consumers should page using `meta.pagination.hasMore` and
-  `meta.pagination.nextCursor`.
+Read [AI agents and the Trekoon skill](docs/ai-agents.md) for installation,
+editor linking, recommended skill combinations, and example prompts.
 
-Payload fields:
+## Read next
 
-```text
-ok: true
-command: task.list
-data:
-  tasks[]: ...
-metadata:
-  contractVersion: "1.0.0"
-  requestId: req-<stable-id>
-meta:
-  pagination: { hasMore, nextCursor }
-```
+- [Quickstart](docs/quickstart.md)
+- [Command reference](docs/commands.md)
+- [AI agents and the Trekoon skill](docs/ai-agents.md)
+- [Machine contracts](docs/machine-contracts.md)
+- [Contributing](CONTRIBUTING.md)
+- [Changelog](CHANGELOG.md)
 
 ## Implementation principles
 
 - Minimal, composable modules
 - Strict validation at command boundaries
-- Stable automation envelope for JSON/TOON modes
+- Stable automation envelopes for JSON and TOON modes
 - No unnecessary feature sprawl