# Build Sequential Multi-Agent Workflows

Complex development tasks often lose context between agents. Stream-Chain passes each step's result forward through custom chains or predefined workflows.

## Install

```bash
npx skillstore add ruvnet/stream-chain
```

## Metadata

- Status: approved
- Slug: ruvnet-stream-chain
- Skillstore revision: r2
- Version status: missing
- Tree hash: 6216b175e186643bc4296fe80c51149c6bbee440f3c5f121a7612b232f1e0619
- Author: ruvnet
- GitHub username: ruvnet
- License: MIT
- Repository: https://github.com/ruvnet/claude-flow/tree/main/.claude/skills/stream-chain
- Ref: ebdfe608f5de2b66ff37ab4af12af8ac4f5e8006
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands
- Quality score: 68
- Public page: https://skillstore.pages.dev/skills/ruvnet-stream-chain
- Manifest: https://skillstore.pages.dev/api/skills/ruvnet-stream-chain/manifest

## Capabilities

- Runs custom chains containing two or more ordered prompts.
- Executes predefined analysis, refactor, test, and optimize pipelines.
- Passes each step's complete output into the next step.
- Supports timeout, verbose, and debug options.
- Defines reusable custom pipelines in the Claude Flow configuration.
- Supports staged code, security, migration, and data-processing workflows.

## Use Cases

- Develop Features in Verified Stages: Move from requirements through implementation, tests, and final review while preserving context between stages.
- Structure Security Reviews: Analyze vulnerabilities, rank findings, propose fixes, and produce security test cases in one ordered workflow.
- Transform and Validate Data: Extract, normalize, validate, and summarize data through separate steps with explicit handoffs.

## Prompt Templates

### Create a Basic Chain

```
Create a Stream-Chain command for [goal] with three prompts: analyze the task, complete the work, and verify the result.
```

### Choose a Predefined Pipeline

```
Select the best predefined pipeline for [repository goal]. Explain the choice and provide a command with an appropriate timeout and logging level.
```

### Design a Reusable Pipeline

```
Design a reusable Stream-Chain pipeline for [workflow]. Define ordered prompts, timeout values, expected handoffs, and a final verification step.
```

### Secure an Advanced Chain

```
Design a Stream-Chain workflow for [complex task]. Separate untrusted data, redact sensitive context, require approval before changes, and include recovery checks.
```

## Limitations

- Requires an installed and configured Claude Flow command-line environment.
- Complete output forwarding can carry unsafe instructions or sensitive data downstream.
- Automatic context persistence can retain sensitive material.
- The included file documents workflows but does not provide their executable implementation.

## Best Practices

- Use specific prompts and order them so each step has a clear input and outcome.
- Add verification and human approval before file changes or command execution.
- Limit forwarded context, redact secrets, and choose timeouts appropriate for each step.

## Anti Patterns

- Do not pass untrusted content into later prompts without clear data boundaries.
- Do not enable verbose or debug logging when prompts may contain secrets.
- Do not chain implementation steps without tests, rollback guidance, or approval gates.

## Security Audit

- Audited at: 2026-07-23T19:30:20.348\+00:00
- Summary: All 87 static alerts are false positives caused by Markdown backticks and fenced examples; SKILL.md contains no Ruby backtick execution or executable implementation. The documented design forwards complete outputs into later prompts, creating an indirect prompt-injection boundary. It also persists chain context automatically, which can retain sensitive material.

## Stats

- Views: 322
- Downloads: 9
- Favorites: 0
- Popularity score: 0