# Unfold

> Understand, diagnose, change, or review an unfamiliar codebase through one evidence-backed mission. Use when learning architecture, tracing a feature, investigating a bug, implementing a real change with guided or agent execution, reviewing a diff or PR, or checking whether tests and the built artifact prove the intended behavior. Prefer over prose-only repository explanations and ungrounded debu…

- **Type:** Skill
- **Install:** `agentstack add skill-railly-skills-unfold`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [Railly](https://agentstack.voostack.com/s/railly)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [Railly](https://github.com/Railly)
- **Source:** https://github.com/Railly/skills/tree/main/foundry/deprecated/unfold

## Install

```sh
agentstack add skill-railly-skills-unfold
```

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

## About

# Unfold

Turn one maintenance mission into a navigable mental model, an evidence-backed change, and proof proportional to the work. Reuse what has already been established instead of restarting the repository tour at every phase.

## 1. Establish the mission

Frame one concrete question or outcome:

> How does `[operation or symptom]` travel from `[entry]` to `[outcome]`, and what evidence would prove the result?

Read existing `.unfold/` artifacts before broad exploration. Compare their recorded commit with HEAD and reverify decisive anchors against current source.

Read [references/protocol.md](references/protocol.md) for the shared evidence and artifact contract.

## 2. Select the earliest incomplete mode

Choose one primary mode. Load only its reference.

| User state | Mode | Reference |
|---|---|---|
| Wants to understand a system or flow | Learn | [learn.md](references/learn.md) |
| Has a symptom, failing behavior, or issue | Triage | [triage.md](references/triage.md) |
| Has a known change to implement | Change | [change.md](references/change.md) |
| Has a diff, patch, test, or PR to assess | Review | [review.md](references/review.md) |

Enter directly at the earliest unfinished state. Do not repeat completed work. A verified Triage Change Surface can enter Change; a finished patch can enter Review; Learn can support any mode when the user lacks a load-bearing part of the mental model.

## 3. Preserve one evidence chain

Carry the same mission, commit, claims, anchors, unknowns, and artifacts across mode transitions. Keep distinctions that change the causal story: setup/runtime/teardown, automatic/manual, caller/owner, persisted/cached, current/proposed, and component/file/function/contract.

When a visual materially clarifies the mission, read [references/visuals.md](references/visuals.md) and render the smallest useful diagram. When durable output is authorized, follow [references/artifacts.md](references/artifacts.md).

## 4. Respect the collaboration policy

Learn is reconstructive by default. Triage remains read-only until the user requests a patch. Change explicitly selects `guided`, `execute`, or `execute-with-approval`. Review may alter tests only when the user authorized review changes.

Never let a mode transition silently broaden write, commit, push, tracker, or release authority.

## Close the mission

Report the result, decisive evidence, remaining unknowns, and the next unfinished mode. The outcome is complete only when the claim is understandable and its proof is retrievable.

## Source & license

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

- **Author:** [Railly](https://github.com/Railly)
- **Source:** [Railly/skills](https://github.com/Railly/skills)
- **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-railly-skills-unfold
- Seller: https://agentstack.voostack.com/s/railly
- 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%.