Free skills — MIT
Ten skills. One file each. No account.
Single-file SKILL.md bundles for Claude Code. Copy one into your project and Claude picks it up from its description. No signup, no API key, no install step, no external process.
The ten skills
Agent Prompt Architect
agent-prompt-architect
Architects the system prompt, context budget, and tool contract of an AI agent as one system.
Context Cartographer
context-cartographer
Assembles high-signal repository context before non-trivial implementation, debugging, or review work.
Context Drift Auditor
context-drift-auditor
Audits whether a prompt or agent run has drifted from its original contract across turns, retrieved sources, edits, and decisions.
Empirical Diagnostician
empirical-diagnostician
Evidence-based debugging: log extraction first, a hypothesis matrix for anything complex, and a root-cause contract before any edit.
Eval Suite Auditor
eval-suite-auditor
Audits an existing evaluation pipeline — LLM judges, assertions, scorers, gates — for silent failure by probing it, not by reading the code.
Prompt Complexity Triage
prompt-complexity-triage
Decides how much a prompt needs to change before rewriting it, then applies only the moves for that tier.
Prompt Evaluation Engineer
prompt-evaluation-engineer
Turns prompts into reproducible evaluation protocols: contracts, test matrices, deterministic checks before rubrics.
Prompt Injection Guard
prompt-injection-guard
A structured protocol for detecting and responding to prompt injection in LLM applications.
Prompt Optimizer
prompt-optimizer
The platform's optimization methodology: classify the prompt, assess sophistication, optimize, and preserve technical parameters.
Subagent Dispatch Economics
subagent-dispatch-economics
Decides whether delegating work to a subagent is worth its cost, and picks the right agent shape.
Install
01
Copy the file
Open the skill’s folder on GitHub and copy its SKILL.md. Each skill is one file — there is no package to install.
02
Drop it in .claude/skills
Create a folder named after the skill and put SKILL.md inside it. Project scope loads it for this repo; home scope loads it everywhere.
03
Ask Claude for that kind of work
Claude reads the skill's description and loads it when your request matches. No command, no registration, no account.
# this project
.claude/skills/<skill-name>/SKILL.md
# every project
~/.claude/skills/<skill-name>/SKILL.mdWhat they are not
A free artifact, with the ceiling stated up front.
These skills are a weaker sibling to the hosted optimizer, not a replacement. Here is exactly where they stop.
- No LLM-based optimization tier — these are instructions Claude follows, not a rewrite pipeline.
- No persistent history, quota tracking, or saved templates.
- No backend call and no key, so nothing is saved anywhere and nothing is metered.
Two different things called skill
Free
The ten above
MIT-licensed files maintained in the public repo. Copy, edit, or fork them. No account and nothing metered.
Generated
A skill package
The Context Engineer turns your goal into a full bundle — SKILL.md plus reference, examples, and helper scripts. Free includes one build a month; Pro includes three.
Want the optimization to run as a server instead of instructions? The cloud MCP package is the supported path — the Free plan includes 20 optimizations a month with no card, though it does need an API key from a free account. The local rules package is deprecated and no longer sold or developed.
Start with the one you need today
Ten files, no account. Add the optimizer later if you want the pipeline behind them.