Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →leeornahum/skill-forge
Create, design, refine, and validate Agent Skills that follow the open SKILL.md standard.
Passed 20 of 23 test cases with the skill, and 9 of 23 without it.
and 15 more in the full report
13 fixed8 passed either way2 still failing1 broke
Tokens, turns and time describe resource use, not a dollar price.
This benchmark measured an earlier version of the skill.
A newer attempt on gemini-3.6-flash on 9/30/2026 did not produce a publishable comparison (not comparable). The snapshot above is the previously measured run.
One of 13 test cases the skill fixed. Same model, same prompt, with and without the skill.
Draft the YAML frontmatter for a skill named `git-rebase-helper` with description `Interactive helper for git rebasing`. The author is `dev-team` and license is `MIT`. Ensure YAML formatting is fully compliant with strict frontmatter parsing requirements.
| Checked | Without skill | With skill |
|---|---|---|
| Encloses all string scalar values in frontmatter in double or single quotes. | Missed | Passed |
| Leaves top-level YAML keys unquoted. | Passed | Passed |
The instructions your model reads when the skill loads.
Create skills that are concise, precisely scoped, and free of context bloat. A skill transfers knowledge the agent lacks, not knowledge it already has.
The upstream source of truth is agentskills.io. The specification and its authoring guides are vendored under references/. This file is the opinionated distillation. The references are the upstream originals, and AGENTS.md owns how they are kept current.
scripts/ directory.Where a reference conflicts with this file on house style, this file wins. The specification still wins on the format contract.
Agents routinely blur these boundaries. Check every skill edit against this table:
| Part | Job | Consumed |
|---|---|---|
name + description | Selection only. They decide whether the skill loads and are never the rules | Read at startup for every skill |
SKILL.md body | The always-needed rules and routing | Loaded when the skill triggers |
references/ | Conditional rules and guidance | Loaded when a routing condition matches |
assets/ | Copyable artifacts | Copied or filled in, not read as guidance |
scripts/ | Executable helpers | Run, not read as guidance |
AGENTS.md | Maintenance contract for editing the skill | Read when editing the skill, never at use time |
README.md | Human skim layer | Read by humans, not loaded by agents |
An instruction that lives only in the description is invisible at execution time, because the description's job ends once the skill loads. Put rules in the body or a reference. In the other direction, procedural detail in the description wastes the selection budget and blurs the trigger.
---
name: "<domain>"
description: "<what the skill does and when to use it>"
license: "<license name or bundled license file>"
compatibility: "<specific environment requirements, when needed>"
test cases passed, vs 9 / 23 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.