code-genesis-fabric

Ready-to-use pipeline agents built on top of the Code-Genesis engine.

Code-Genesis is the AI pipeline engine that powers everything here — it handles orchestration, parallel execution, agent contexts, and step control. This repo is purely a collection of pipelines and agents you can drop in and run.

---

What's inside

code-genesis-fabric/
├── samples/                    # Learning pipelines — start here
│   ├── hello-world.yml         # Minimal 3-step pipeline (plan → build → review)
│   ├── fail-if-step.yml        # Guard rails: stop the pipeline on LLM-detected failure
│   ├── foreach-parallel.yml    # Nested foreach + parallel branches per item
│   ├── parallel-foreach.yml    # Concurrent iteration over a collection
│   └── contexts/               # Reusable agent context definitions
│       └── planner/
│
└── snyk-issues-report/         # Production-ready Snyk security report pipeline
    ├── snyk-issues-report.yml  # Discovers tagged projects → fetches issues → exports CSV
    └── contexts/               # Agent contexts used by this pipeline
        ├── snyk-discoverer/
        ├── snyk-validator/
        ├── snyk-issue-fetcher/
        ├── snyk-csv-builder/
        └── snyk-file-writer/

---

Samples

Minimal pipelines that demonstrate the core Code-Genesis primitives.

File	What it shows
hello-world.yml	Basic sequential steps with context, prompt, and output_key
fail-if-step.yml	Using fail_if + fail_message so the LLM can abort the pipeline
foreach-parallel.yml	foreach iteration with nested parallel branches inside each item
parallel-foreach.yml	parallel_foreach with max_concurrency and fail_fast

---

Pipelines

snyk-issues-report

Generates a CSV report of open critical and high severity vulnerabilities across all Snyk projects that share a given tag.

Required MCP: mcp-snyks

This pipeline will not work without the mcp-snyks MCP server running. All Snyk API calls are made exclusively through it. Register it in your Claude MCP configuration using HTTP transport:

{
  "mcpServers": {
    "mcp-snyks": {
      "type": "http",
      "url": "http://localhost:<PORT>/mcp"
    }
  }
}

See mcp-snyks for setup instructions, available port, and required environment variables (Snyk API token, etc.).

How it works:

[1] Discover tagged projects      — finds all projects matching tag_key + tag_value
[2] Validate project list         — aborts early if no projects are found (fail_if guard)
[3] Fetch issues (parallel)       — parallel_foreach across all projects (10 concurrent)
[4] Build CSV report              — flattens, categorizes, and sorts all issues
[5] Save to disk                  — writes the final CSV file

Inputs:

Input	Description	Default
org	Snyk organization name, slug, or GUID	my-org
tag_key	Tag key to filter projects	team
tag_value	Tag value to match	backend
output_file	Destination path for the CSV	snyk-issues-report.csv

Outputs:

Output	Description
report_saved	Confirmation with issue counts and saved file path
csv_content	Full CSV text (header + rows + summary block)
raw_issues	Raw per-project issue arrays before aggregation
projects_found	Projects that matched the tag filter

---

Related

• code-genesis — the engine that runs these pipelines