L-GEVITY skills (ALCHEMY)
26 engineering skills with a router: the agent picks methods, forms a team and checks results against evidence.
- 81 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Worker agents write to the repository, run checks and prepare a pull request that waits for your approval.
- The installer writes skill files and root instructions into the project.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/l-gevity-skills-alchemyDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Worker agents write to the repository, run checks and prepare a pull request that waits for your approval.
- The installer writes skill files and root instructions into the project.
This is third-party code. Review the repository files before installing.
What it does
ALCHEMY takes an objective and decides which engineering questions it raises: whether a feature earns its complexity, where a module boundary belongs, how to verify it, what limits delivery. The `alchemy` router selects the questions, and `alchemy-orchestrate` assigns work, orders dependencies and checks hand-offs. One agent is the default. Independent parts run in parallel, and risky changes get a separate reviewer. Completion is backed by executed checks and an independent review, and the prepared pull request waits for your approval. Invoke it with `/alchemy` in Claude Code or `$alchemy` in Codex.
Who it is for. Engineers and tech leads running a refactor or a large feature through an agent who want a verifiable process instead of one long session.
Good fit when
- Before a big feature you want to decide if it is worth the complexity, so you call a single focused gate.
- An auth refactor is split across several agents and write boundaries need watching.
- You ask for a structural audit of a pull request without traversing the whole repository.
- A recurring agent mistake should become a skill improvement.
Not a fit when
- You need a quick local edit with no analysis and a plain session is enough.
- Skills need framework-specific guidance that the set lacks and you are not ready to add your own.
- The environment does not let agents run the Python check scripts.
Example request
Do some alchemy on this auth refactor: first decide whether the functionality is needed, then check module boundaries.Limitations
The library covers structural reasoning: requirements, architecture, verification strategy and delivery reliability. It has no knowledge of your framework, security, accessibility or release policy, so you add those with your own skills. Agent-team mode is enabled separately with the `init_workspace.py` script for Claude Code, Codex or Unstoppable, and support depends on the runtime. Installers exist for Claude Code, Codex, Gemini CLI and Grok CLI. An existing root instruction file is never overwritten: a copy with a `.l-gevity` suffix lands beside it for manual merging.
How to disable. Delete the skill folders from `.claude/skills/` or `.agents/skills/`, the `l-gevity-skills.lock.json` file and the added root instruction files. Remove the `alchemy-workspace.json` profile and the `.alchemy/runs` directory if you enabled the team.
FAQ
How do I invoke ALCHEMY?
In Claude Code type `/alchemy` plus the task, in Codex `$alchemy`. The phrase "do some alchemy" works too. `/alchemy ?` shows help and `/alchemy audit this PR` starts a structural audit.
What are the seven gates?
The order is M, A, L, C, E, H, Y: minimum, architecture, locality, complexity, enforcement, hermetic and yield. Value is checked before design and optimization.
How many agents run by default?
One. The `execution.max_parallel_agents` key defaults to 3 and `execution.max_rework_cycles` to 2. Setting `execution.multi_agent` to `off` keeps everything in one session.
How do I pin the skills version?
Set `L_GEVITY_SKILLS_REF` to a branch, tag or commit. File provenance is recorded in `l-gevity-skills.lock.json`.
Which agents have installers?
Claude Code, Codex, Gemini CLI and Grok CLI on Windows, Linux and macOS. Skills go to `.claude/skills/` or `.agents/skills/`, and the root instruction file is written beside yours.
Discussion
No questions or reviews yet.
Related
SkillsOver twenty skills for coding agents: MCP and hooks, plan mode, a bug-fix protocol, prompting and a bash safety hook
14 skills that find the gaps in a task before the code: interviews, plans, assumption tests and proof that a bug fix works
Claude Code kit with eight phase agents, 21 skills and hooks that keep the agent inside set limits.
39 engineering-discipline skills and 5 subagents for Claude Code: root-cause debugging, security, edge cases, honest reporting
A 2.4 KB charter for CLAUDE.md and three on-demand skills: the agent stops guessing, verifies results and gates irreversible steps
Six commands for Claude Code, Codex and Cursor: the main session holds context while fresh agents review the problem and diff.
README in short
Skills install through a prompt to your agent or ready-made installers, and the agent team is enabled by a separate script. Results are recorded in plan, hand-off and verification ledger files.