Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+22 more

    Make Skill and Publish

    1

    Turn a plain-language idea into a validated, installable Agent Skill - no spec reading required. Includes a pre-publish scan for secrets and PII.

    Free

    3 installsSecurity scanned
    Make Skill and Publish

    Make Skill and Publish

    Example session with this skill installed

    Turn my SillyTavern install into a skill: I want my agent to start it, stop it, tell me which port it is on, change settings safely without breaking them, and diagnose it when it will not start. Then publish that skill.

    • Read your context and instructions
    • Compiled the make and publish

    Summary

    Built sillytavern-controller from that one-line requirement: scoped, scaffolded, written, tested against the real installation, then scanned and validated. Validation passes for both skills in the repo and the pre-publish scan is clean.

    Pre-publish checkResult
    validate_skill.py[OK] skills/make-skill-and-publish - [OK] skills/sillytavern-controller
    scan_publish_ready.py20 files scanned - HIGH=0 MEDIUM=0 LOW=0
    Trigger test (10 should-fire + 10 near-miss)not executed - no second agent session was available, recorded as unresolved rather than claimed
    Findings raised and fixed1 - a masked-credential example tripped the secret scanner, rewritten and re-scanned clean

    Live test round against the actual install:

    root located   server.js + config.yaml + src/server-main.js all present
version        1.19.0
config         port: 8000   listen: false   whitelistMode: true   dataRoot: ./data
port owner     nothing listening on 8000
http probe     000 - connection refused
verdict        the server is not running. Not a port conflict, not a whitelist block.
next action    start it, or leave it down - your call

    skills/
`-- sillytavern-controller/

    |-- SKILL.md
    |-- agents/openai.yaml
    |-- references/
    |   |-- operations.md
    |   `-- troubleshooting.md

    `-- assets/trigger-eval-queries.json

    Next steps

    • Read the scan output above and approve the commit, or tell me what to change first
    • Nothing is committed or published yet - this run stops at the approval gate by design

    Connects securely to your tools. The creator never sees your data.

    About this skill

    What it does

    Takes whatever you already have - a sentence, a document, a screenshot, an existing workflow - and drives it all the way to a finished Agent Skill: scoped, scaffolded, written, self-tested, scanned, committed and published to GitHub as something anyone can install.

    Why you need this

    Writing a SKILL.md is easy. Getting it right is where people lose hours:

    • name has to match the folder, be lowercase-hyphen, and stay under 64 characters
    • description is the routing surface - it decides whether your skill ever fires - and has to be written as "Use when ..."
    • The body must stay lean; long procedures belong in references/, each linked with the condition for reading it
    • The pre-publish pass is the part nobody does well: secrets, tokens, private keys, PII, machine-specific absolute paths, pycache, .DS_Store, stray test debris

    This skill encodes all of that as a fixed workflow, so the standard gets applied every time instead of being remembered every time.

    Built for people with ideas, not just people who know the spec

    The people who need this most are not the ones who already know the format. They are the ones with the ideas.

    If you can describe in plain language what you want your agent to do, but you get stuck on the mechanics - folder layout, frontmatter rules, how to phrase a description so it actually fires, what to test, what to clean up before publishing - that is exactly the part this skill takes over.

    You bring the idea. The skill handles the spec.

    And that is the point. The barrier is almost never the idea; it is the process wrapped around it. When finishing a skill stops costing a week of reading specification documents, people actually finish them. The first one lands, and the rest come faster. Take the busywork out of publishing, and ideas stop piling up in notes and start shipping.

    How it works

    Nine steps, with exactly one hard gate:

    1. Turn the request into a skill spec - and say what is missing instead of padding it
    2. Scaffold the skill folder
    3. Write SKILL.md and its resources
    4. Test against a realistic request, and mark a test "not executed" rather than implying it passed
    5. Write the repo README
    6. Run the pre-publish scan and report the raw output
    7. STOP - the human authorizes
    8. Publish to GitHub
    9. Verify and report

    Safety by design

    • Never handles credentials. It will not ask for, echo or store passwords, tokens or 2FA codes. Authentication stays with you.
    • One explicit gate. git add, commit, creating a remote and any upload all wait until you have seen the scan output and said go.
    • Never fabricates. Descriptions, README and examples must describe what the skill actually does.
    • Portable output. Relative paths only - no machine-specific absolute paths, usernames or local test debris in the published tree.

    What is inside

    • SKILL.md - the workflow, hard rules and gotchas
    • references/stage-prompts.md - ready-to-use prompt templates per stage, including a one-shot version
    • references/publish-github.md - installing and authenticating gh, creating the repo, publishing a release, troubleshooting
    • scripts/new_skill.py - self-contained scaffolder, standard library only
    • scripts/validate_skill.py - spec validator
    • scripts/scan_publish_ready.py - the pre-publish scan, usable as a CI gate
    • assets/trigger-eval-queries.json - a worked trigger-test set: 20 labelled queries with observed outcomes

    How to use

    Skills on Agensi run inside your AI agent. Copy the prompt, paste it into your agent, and it connects to Agensi and installs this skill for you. If a payment is needed, your agent shows it to you first.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill passes our trust checks below.

    Security scanned

    Passed our 8-point scan before listing

    3 installs

    Downloaded by developers to date

    Free forever

    No account required to browse

    Trust & safety

    Security scanned

    Verified clean 7 days ago

    • Free to download with an account

    Requires

    GitHub

    Listed7 days ago

    What's inside

    Frequently Asked Questions