# Bundled Skill File Extraction

> Lazily extract referenced files for bundled skills so large prompts stay offline until needed.

- **Type:** Skill
- **Install:** `agentstack add skill-ychampion-cskill-agents-bundled-skill-file-extraction`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [ychampion](https://agentstack.voostack.com/s/ychampion)
- **Installs:** 0
- **Category:** [Agent Skills](https://agentstack.voostack.com/c/agent-skills)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [ychampion](https://github.com/ychampion)
- **Source:** https://github.com/ychampion/cskill-agents/tree/main/agents/claude-code/skills/bundled-skill-file-extraction
- **Website:** https://x.com/YashasGunderia/status/2039990540016693692?s=20

## Install

```sh
agentstack add skill-ychampion-cskill-agents-bundled-skill-file-extraction
```

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

## About

# SKILL: Bundled Skill File Extraction
**Domain:** command-surfaces  
**Trigger:** Use whenever a bundled skill needs reference files extracted on demand while keeping startup fast.
**Source Pattern:** Distilled from reviewed command-surface and listing-control patterns.

## Core Method
When a bundled skill defines `files`, memoize an extraction promise and write each file under a secure directory (with O_NOFOLLOW / 0o700+ permissions). Cache the extraction path so the same directory can be reused across invocations, and wrap the skill prompt with `Base directory for this skill: ` so the agent can Read/Grep as needed without bloating the inline instructions.

## Key Rules
- Group requested files by parent directory when writing so mkdir/disk writes happen per subtree.
- Allocate per-process directories under the locked `getBundledSkillsRoot()` and avoid overwriting by using `O_EXCL`/`SAFE_WRITE_FLAGS`.
- Memoize the extraction promise (not just the result) so concurrent invocations await a single write.
- Prepend `Base directory ...` to the skill prompt only when the extraction succeeded; otherwise continue without the prefix to keep the skill usable.

## Example Application
When packaging a bundled “project review” skill with large README snippets, add `files` entries for the README, let this method extract them lazily, and cite the base directory in the prompt so the agent can `Read` the files only when it actually runs the skill.

## Anti-Patterns (What NOT to do)
- Don’t write the files directly during startup—delay until the first invocation, using memoized promises to avoid race conditions.
- Don’t forget to guard against traversal (`resolveSkillFilePath`) so bad relpaths cannot escape the extraction dir.

## Source & license

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

- **Author:** [ychampion](https://github.com/ychampion)
- **Source:** [ychampion/cskill-agents](https://github.com/ychampion/cskill-agents)
- **License:** MIT
- **Homepage:** https://x.com/YashasGunderia/status/2039990540016693692?s=20

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-ychampion-cskill-agents-bundled-skill-file-extraction
- Seller: https://agentstack.voostack.com/s/ychampion
- 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%.