Index of every page on this site: llms.txt
All skills
milind-soni avatar

/create-verification-skill

@3306c52
by Milind Sonimilind-soni/openmausbot26k stars
3,644
Run it yourself

Create one skill that teaches a bot how to launch, drive, and verify a specific app or project. Use only when the user explicitly asks for a verification skill.

Use this Skill: https://skilld.dev/gh/milind-soni/openmausbot/create-verification-skill

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈47 tokens for metadata: the name and description. ≈708 when used: this file. ≈95 more on demand in 1 file.

Description uses 2.4% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈188 characters, or 2.4%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Create a verification skill

Create one compact, project-specific SKILL.md through skill_manage, which shows the person every change with an Undo. Do not install skill files directly or silently add scripts to the user's project.

1. Inspect before drafting

Use the tools already mounted on this bot to establish:

  • the exact project or app being verified;
  • its existing launch command and reliable ready signal;
  • the safest available control surface: existing project CLI, browser, computer, phone, or shell;
  • one read-only health check;
  • up to three important user workflows and the observable result that proves each one worked;
  • how to stop only the process or resource started by the verification run.

Prefer an existing control surface. If the project has no dependable way to launch or drive the relevant workflow, report that prerequisite instead of inventing commands, selectors, APIs, or a new control program.

2. Prove one workflow

Run the real launch, health check, one representative workflow, evidence capture, and cleanup once. Use an isolated test profile or temporary data when available; never drive the user's live data merely to author the skill. If the workflow cannot be proved with the tools currently mounted, stop and explain which capability is missing.

3. Draft one self-contained skill

The proposed SKILL.md must contain YAML frontmatter and these sections:

  1. Launch — exact command, isolated data/profile, ready signal.
  2. Doctor — one read-only check with an actionable failure message.
  3. Drive — stable names, accessibility targets, routes, refs, or commands; never recorded screen coordinates.
  4. Evidence — action plus resulting state and any important side effect.
  5. Cleanup — stop only what this run started; preserve the evidence.
  6. Feature map — at most three proven workflows, each with the user path, control recipe, success proof, and known gotcha.
  7. Maintenance — re-run Doctor and the affected workflow when the app changes; update drifted instructions but never hide a product failure.

Keep it concise. Evidence about the user path is input to the draft—not permission to retain secrets, audio, screenshots, or coordinates.

4. Save it

Call skills_list first. If an existing skill already covers the project, report that and do not overwrite it. Otherwise call skill_manage with action="create", the complete skill_md, a short gist, and the exact source used (repository path, URL, or conversation).

Read the skill_manage result. If it applied, tell the user the skill's name, that it is on, and that they can undo it from the chat. Only if it is pending, tell them it stays inactive until they decide on the in-app card. Never claim it was enabled without an applied result.

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No alerts18d3 checks · Risk SAFE
  • Gen Agent Trust Hub18d

    This skill provides a structured way for the agent to create verification tests for projects. It includes strong safety guidelines, such as using isolated test data and avoiding the silent installation of scripts, though it inherently involves executing local project commands and reading project configuration files.

  • Socket18d

    No alerts

  • Snyk18d

    Risk: LOW · No issues

Signed by skilld at 3306c52. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated last month

README badge

README badge for milind-soni/openmausbot/create-verification-skill