homerail-dag-ops
Design, start, supervise, inspect, and continue HomeRail DAG workflows from any agent with CLI or HTTP access.
Skill by xiaotianfotos
Chalk Labs
Need help deploying homerail-dag-ops?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/xiaotianfotos-homerail-dag-ops
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: design, productivity, agent-building
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Python package
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/homerail-dag-ops
What's inside
- Skill name
homerail-dag-ops- Description
- Design, start, supervise, inspect, and continue HomeRail DAG workflows from any agent with CLI or HTTP access. Use for project work delegated to a DAG, workflow patterns, existing-run results, approval/Actor follow-ups, live panels, and background event callbacks. Handles the full run lifecycle with event-driven waiting instead of repeated model polling.
- License
- MIT
- Instructions
- 98 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (15)
- .gitignore
- SKILL.md
- openai.yaml · agents/
- acceptance.md · references/
- architecture.md · references/
- inspection.md · references/
- listener.md · references/
- multi-actor-surfaces.md · references/
- patterns.md · references/
- supervision.md · references/
- check_integration.py · scripts/
- dag_subscription.py · scripts/
- manager_fixture.mjs · scripts/
- storage.py · scripts/
- test_dag_subscription.py · scripts/
SKILL.md (first lines)
# HomeRail DAG Use one workflow: **define outputs → choose/validate a DAG → start once → wait outside the model → handle decisions → verify evidence**. Running and supervising are stages of the same task. Load only the reference needed for the current stage; do not load every HomeRail skill before starting. ## Choose the next action | Request or current state | Action | | --- | --- | | New project work | Use the start path below; reuse a workflow/profile. | | New topology or a reusable pattern | Read [patterns.md](references/patterns.md); inspect the live schema/catalog. | | Active run requiring event-driven waiting | Read [supervision.md](references/supervision.md); use the current host’s blocking event tool or a verified optional adapter. | | `[homerail-dag-event ...]` callback, including the former supervision skill name | Follow event consumption in [supervision.md](references/supervision.md). | | Requested status, results, or diagnosis | Read [inspection.md](references/inspection.md); inspect only relevant evidence. | | Live Actor panels or a later per-Actor command | Read [multi-actor-surfaces.md](references/multi-actor-surfaces.md). | The old `homerail-dag-patterns` entry remains for compatibility. Its content lives in this skill's references; it is not a prerequisite skill to load. ## Start once 1. Establish the output, source revision/inputs, permitted changes, budget, and acceptance evidence. If unattended supervision is needed, confirm the host can await a tool/event or receive a callback before starting the run. If a run ID already exists, inspect that run first. 2. Resolve the existing CLI and Manager origin. Use `hr doctor` for readiness; use `hr runtime status` if diagnosis is needed. If `hr` is absent but `HOMERAIL_CLI_ENTRYPOINT` is set, use `node "$HOMERAIL_CLI_ENTRYPOINT"`. Do not rebuild or restart a ready installation for an ordinary DAG run. Use `homerail-install-ops` only for an actual installation/service task. 3. Use `hr --json templates list`, or inspect a known database workflow. Read its input contract. Resolve template files from the verified HomeRail source or Manager catalog; `assets/...` is not relative to an arbitrary project. 4. Give Workers reachable inputs. Container Workers do not automatically see host paths: supply a cloneable repository URL and pinned revision, or the workflow's supported workspace/input mechanism. Keep input identity shared across nodes and verify access from the executing environment. 5. For a new/changed workflow, inspect `hr dag schema`, validate with `hr --json dag validate <file>`, then sync. Bind database LLM settings through a runtime profile; keep provider/model credentials out of workflow YAML. ```bash # Existing database workflow and runtime profile hr run --workflow <workflow-id> --profile <profile-id> --prompt "<task and inputs>" # Alternatively: a concrete template with its actual input contract hr run <verified-template-path> --prompt "<task and inputs>" ``` Template-embedded profiles and database profiles differ: use `--workflow` for a DB profile or the documented sync flow. Check local help for unfamiliar flags. Record the returned run ID, workflow revision/hash, inputs and acceptance criteria in a private receipt. If submission has an unknown outcome, reconcile existing runs; do not blindly create a replacement. In the Manager Agent, available tools such as `list_orchestrations`, `create_and_run`, `invoke_run`, `get_run_status`, `get_dag_schema`, and `validate_dag_workflow` are shortcuts for covered operations. Use structured CLI output for the rest. Any agent may use these CLI/HTTP contracts; no particular supervisor model, SDK or harness is required. Do not launch another agent or model merely to subscribe to or consume events. ## Wait, decide, and finish The default bundled path is `register` → `wait` → handle the returned event → `ack`. Registration persists identity; `wait` blocks in ordinary code and returns one actionable JSON event. Let the host await that tool instead of repeatedly invoking the model. Registration alone does not start a background observer. If the host provides native event subscriptions, use its equivalent mechanism. An optional callback adapter can deliver the same JSON into a persistent host session. End the model turn after registration only when that continuing event path is verified. A plain skill cannot force a host to suspend, steer or resume an agent. If the host lacks event/tool continuation, explain that limitation; do not claim automatic wakeup or substitute a Codex executor. Routine progress stays quiet. `hr dag watch`/`supervise` remain optional foreground inspection tools, not a model polling loop. A requested snapshot needs no new watcher. On a callback, verify the single event and current run identity, handle the current condition, record the finding/action intent, and acknowledge consumption. An ACK does not approve a proposal or end a waiting DAG. Let the human decide approval nodes. Reconcile ambiguous mutation outcomes before retrying. Check terminal status and relevant artifacts/handoffs against the original acceptance criteria. A completed pipeline or passing advisory scorecard alone is not proof of useful output. See [inspection.md](references/inspection.md). End the turn when nothing actionable remains. Keep secrets in existing Manager configuration/encrypted storage, never in prompts, templates, receipts or commits. Keep deployment and disruptive runtime changes within the user's authorized scope. For HomeRail self-development read [architecture.md](references/architecture.md); for skill changes read [acceptance.md](references/acceptance.md).
From xiaotianfotos/homerail, MIT.
Signals
- Stars
- 998
- Forks
- 226
- Open issues
- 19
- Last commit (this folder)
- Sep 13, 2026 (2 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/xiaotianfotos/homerail
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get homerail-dag-ops running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
homerail-dag-ops skill. Agent Mart. https://agentmart.dev/skills/xiaotianfotos-homerail-dag-ops (last updated 2026-10-24)