# Ticket Work

> Contract-driven ticket execution with Linear integration - orchestrates intake, research, questions, spec, implementation, review, and done phases with explicit human gates

- Skill: `omninode-ai/ticket-work` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add omninode-ai/ticket-work`
- Raw SKILL.md: https://api.skillmd.com/api/skills/omninode-ai/ticket-work/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: OmniNode-ai (https://skillmd.com/u/omninode-ai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/omninode-ai/ticket-work

---


# Contract-Driven Ticket Execution

**Announce at start:** "I'm using the ticket-work skill."

## Usage

```
/ticket-work PROJ-1807
/ticket-work PROJ-1807 --autonomous
/ticket-work PROJ-1807 --skip-to implement
```

## Execution

### Step 1 — Parse arguments

- `ticket_id` → Linear ticket ID (required)
- `--autonomous` → skip human gates (for pipeline/overnight use)
- `--skip-to` → resume from named phase

### Step 2 — Initialize node (contract verification)

```bash
onex run-node node_ticket_work \
  --input '{"ticket_id": "<ticket_id>", "autonomous": false, "skip_to": null}' \
  --timeout 300
```

On non-zero exit, a `SkillRoutingError` JSON envelope is returned — surface it directly, do not produce prose. Note: handler is a structural placeholder; full migration is tracked separately.

### Step 3 — Execute ticket phases via tracker DI

This skill uses `ProtocolProjectTracker` DI (resolved via `resolve_project_tracker()`) for all
Linear operations. The tracker adapter handles routing automatically.

1. **INTAKE**: Fetch ticket via `tracker.get_issue()`; parse contract YAML block
2. **RESEARCH**: Read relevant code, check existing implementations — see Research Enrichment below
3. **QUESTIONS**: Surface blockers → human gate (skipped in `--autonomous` mode)
4. **SPEC**: Write implementation spec to ticket description
5. **IMPLEMENT**: Write code in git worktree; run tests; pre-commit clean
6. **REVIEW**: Run local-review skill; address findings
7. **DONE**: Create PR; update Linear ticket status to In Review

### Step 4 — Report

Write `ModelSkillResult` to `$ONEX_STATE_DIR/skill-results/{context_id}/ticket-work.json`.
Display: phase reached, PR URL, any blocking questions.

## Research Enrichment Flow

Before proceeding from RESEARCH to QUESTIONS, the node dispatches a context request to
`node_ticket_research_enrichment_compute`. This enriches the research phase with assembled
knowledge context based on the ticket's repo, description, and linked files.

```bash
onex run-node node_ticket_research_enrichment_compute \
  --input '{
    "ticket_id": "<ticket_id>",
    "repo": "<repo>",
    "description": "<ticket_description>",
    "linked_files": ["<file1>", "<file2>"],
    "context_timeout_s": 10.0
  }' \
  --timeout 15
```

**Output written to:** `$ONEX_STATE_DIR/knowledge-context/<ticket_id>.md`

**Non-blocking by design:** if the context assembler times out (>10s default) or errors,
research continues without context — enrichment is advisory, not required. The result
status (`ok` | `timeout` | `error` | `skipped`) is logged but does not gate phase progression.

**Enrichment statuses:**

| Status | Meaning |
|--------|---------|
| `ok` | Context assembled and written to audit path |
| `timeout` | Assembler exceeded `context_timeout_s`; research proceeds without context |
| `error` | Assembler raised an exception; research proceeds without context |
| `skipped` | No assembler configured (default; safe to ignore) |

## Architecture

```
SKILL.md   -> thin shell (this file)
node       -> omnimarket/src/omnimarket/nodes/node_ticket_work/ (structural placeholder)
contract   -> node_ticket_work/contract.yaml
enrichment -> omnimarket/src/omnimarket/nodes/node_ticket_research_enrichment_compute/
```