# Seed To Task Graph

> Light recipe: create work/task-graph.md directly from a seed. Eager-gate on sparse seeds before decomposing.

- **Type:** Skill
- **Install:** `agentstack add skill-xmu-csnoob-praxiskit-seed-to-task-graph`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [xmu-csnoob](https://agentstack.voostack.com/s/xmu-csnoob)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [xmu-csnoob](https://github.com/xmu-csnoob)
- **Source:** https://github.com/xmu-csnoob/PraxisKit/tree/main/plugins/praxiskit/skills/seed-to-task-graph

## Install

```sh
agentstack add skill-xmu-csnoob-praxiskit-seed-to-task-graph
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Seed to Task Graph

Produce a task graph directly from a rough seed. **Skips `idea` and `prd` artifacts entirely.** Most seeds are 1-2 sentences — the gate fires eagerly precisely because there's no intermediate idea/PRD to absorb the drift.

```text
seed -> seed-to-task-graph -> work/task-graph.md + work/SUBAGENT.md -> task-graph-to-batch
```

## Contract

**Inputs:** `seed` (raw user input — text, file path, or pasted prompt)
**Output:** `work/task-graph.md` + `work/SUBAGENT.md` per `schemas/task-graph.schema.md` (v3.0); fires `references/clarification-gate.md` on missing goal/acceptance/forbidden-to-infer fields
**Preconditions:** seed (text or file) provided; seed contains at least an implicit goal (or one is collected via gate)
**Stop boundary:** Does NOT execute tasks, write idea.md or PRD.md, or invoke other transforms.

## When to Use This vs `seed-to-idea`

Light recipe (this skill) — script, prototype, refactor, bug fix; user can directly state acceptance criteria; no external stakeholders need an idea/PRD.

Standard recipe (`seed-to-idea` → `idea-to-prd`) — multiple stakeholders, acceptance criteria require derivation from product goals, surface area benefits from explicit non-goals.

The user picks the recipe; this transform does NOT auto-classify.

## Workflow

1. **Load the seed.** If user provides a path, read it. If pasted text, use it.
2. **Load gate reference + schemas.** Read `references/clarification-gate.md`, `schemas/seed.schema.md`, and `schemas/task-graph.schema.md`.
3. **Triage seed length.** A 1-2 sentence seed means most goals/acceptance signals are gaps. Apply the Sparse-Input UX Rule from clarification-gate: gate first, draft after.
4. **Triage seed against `seed.schema.md`.** Identify which triage dimensions are present.
5. **Apply cross-field rules** from `idea.schema.md` (distributed → consistency_model; caching with metadata → cache_metadata_fields). Add triggered fields to gaps.
6. **Required gate fields:**
   - `goal_statement` if not derivable from seed
   - `acceptance_signal` only if the seed does not already provide observable done criteria. If present, copy/quote/split the seed's done criteria into task acceptance criteria and annotate `[user]`.
   - Any forbidden-to-infer field triggered by cross-field rules
7. **Call clarification-gate** with collected gaps. Prefer host-native multi-question UI (group up to 4).
8. **Decompose and write artifacts.** Follow `references/decompose-task-graph.md` (steps 1-10). Frozen contracts step is typically skipped on light recipe.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [xmu-csnoob](https://github.com/xmu-csnoob)
- **Source:** [xmu-csnoob/PraxisKit](https://github.com/xmu-csnoob/PraxisKit)
- **License:** MIT

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/skill-xmu-csnoob-praxiskit-seed-to-task-graph
- Seller: https://agentstack.voostack.com/s/xmu-csnoob
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.