Compile and execute declarative multi-agent pipelines using the Graft CLI runtime. Compose agent graphs, state handoffs, and verification steps into repeatable workflows. Use when orchestrating cooperating agents.
Installs into .claude/skills of the current project.
Sponsors
SSkills DirectorySkills Directory
Know which skills are safe — weekly.
SSkills DirectorySkills Directory
Is that skill safe to install?
Are you the author of Graft?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/siddu1004-graft)
---
name: graft
description: "Compile and execute declarative multi-agent pipelines using the Graft CLI runtime. Compose agent graphs, state handoffs, and verification steps into repeatable workflows. Use when orchestrating cooperating agents."
---
> **This skill needs Graft, which is installed separately.** From this skill's folder, run `python3 scripts/sfp_app.py status` to see whether it is installed and running. If it is not, run `python3 scripts/sfp_app.py install` to get the plan (it changes nothing), tell the user what it will do, and run it again with `--yes` only after they agree. `start` runs it when it is installed but stopped.
# Graft Multi-Agent Pipeline Compiler
Declarative multi-agent orchestration runtime and compiler CLI. Graft lets developers define agent state machines, tool access policies, and validation gates in structured configuration files.
## Use this skill when
- Composing pipelines where multiple specialized agents collaborate sequentially or in parallel.
- Setting explicit handoff schemas and validation gates between agent stages.
- Replaying and debugging multi-step agent execution traces.
## Steps
1. **Initialize a Graft pipeline project**:
Scaffold pipeline configuration:
```bash
graft init my-agent-pipeline
```
2. **Define agent nodes and edges**:
Specify agents, tasks, and transition conditions in `graft.yaml`:
```yaml
pipeline:
- name: researcher
role: "Codebase exploration"
output: "context.json"
- name: implementer
depends_on: [researcher]
role: "Code authoring"
```
3. **Validate pipeline graph**:
Run schema and dependency cycle checks:
```bash
graft check graft.yaml
```
4. **Execute pipeline run**:
Execute the compiled pipeline with live logging:
```bash
graft run graft.yaml --log-level info
```
## Checks
- Verify graph has no circular dependencies or unresolvable transitions.
- Check that output schemas match input contracts of downstream nodes.
## Pitfalls
- Ensure environment variables and API credentials required by subagents are declared in project secrets.
- Always include explicit fallback and timeout transitions on long-running tasks.
## Sources
- [Graft GitHub Repository](https://github.com/JSLEEKR/graft)