agentleFS
Schramm2Schramm2/spyglass/skills/create-verification-skill/SKILL.md

create-verification-skill

Skill0 starsChanged 25 days agoMITView on GitHub

Create a project-owned verification skill and feature map that use Spyglass with the repository's own tools and lifecycle.

92 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use setup-pstack, ketchup-video and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • bnemaSkill

    setup-pstack

    Configure which Pi models pstack uses per role. Detects available provider/model specs and writes pstack's user configuration. Use for /setup-pstack, "configure pstack models", or changing pstack model choices.

    Also in 25 projects that use this

  • tristanphvnSkill

    ketchup-video

    Catch the user up like /keel:ketchup, with a short subtitled video of slides for what happened and the waiting calls as text under it. Use for /keel:ketchup-video.

    Also in 13 projects that use this

  • ianclearySkill

    poteto-help

    Guides users through pstack setup, /poteto-mode, and picking the skill, playbook, or principle for a task. Type /poteto-help with a question.

    Also in 14 projects that use this

  • eryzerzSkill

    make-bot-ui

    Use when building a custom UI (page, dashboard, buttons) that should wake a Grok Bot over a webhook, when the user must provide a webhook sender key, or when exposing that UI on Tailscale.

    Also in 19 projects that use this

  • ianclearySkill

    principle-explain-the-number

    Apply before you trust, report, or act on a number you measured: a speedup, a regression, a throughput, a latency, or an eval result. Find what limits it, and rule out that it measured something other than the work you think.

    Also in 27 projects that use this

  • drafaelSkill

    in-your-own-words

    Restate the user's goals and the problem they are trying to solve to check understanding. Manual invocation only; use when the user explicitly invokes in-your-own-words or requests this understanding check. Do not solve the task or rewrite the assistant's last answer.

    Also in 26 projects that use this

The file

Show its rules and text

We split this skill into the rules it gives an agent

20 rules. 0 are common, 10 are distinctive, and 10 are very specific to this project.

  • 0

    Common

    Most of the 79 edited versions we read keep these.

  • 10

    Distinctive

    This author's own choices. Worth a look.

  • 10

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Create a project-owned verify-<app> skill so the next agent can verify the app from a cold start.

    line 9
  2. Distinctive:

    Keep the repository's own commands in charge of its installations, databases, and services.

    line 11
  3. Distinctive:

    Read repository instructions, README, manifests, task runners, existing verification skills, and feature maps.

    line 16
  4. Distinctive:

    If the repository already has a verification skill, preserve its proven journeys and owner commands.

    line 35
  5. Distinctive:

    If the app cannot start or the owner check fails, record the concrete failure.

    line 37
  6. Very specific:

    Read the installed CLI reference and schemas: spyglass --help, spyglass schema profile, and spyglass schema plan.

    line 42
  7. Very specific:

    If Spyglass is not installed, follow the Spyglass README's terminal setup, then check spyglass --version before creating the profile.

    line 49
  8. Very specific:

    Run spyglass init when the repository has no profile.

    line 52
  9. Very specific:

    Keep the repository owner in control of setup and cleanup.

    line 57
  10. Very specific:

    Use exec for a trusted owner command that has no named check.

    line 60
  11. Distinctive:

    Write a claim plan before driving.

    line 63
  12. Very specific:

    Write SKILL.md with name: verify-<app> and a description that names the app, its main surface, and when to use the skill.

    line 74
  13. Very specific:

    Create features/README.md and files for the top three to five user-facing features, selected from routes, commands, menus, or library …

    line 79
  14. Very specific:

    Keep changing command syntax in Spyglass help and profile schemas.

    line 93
  15. Distinctive:

    Before handoff, use the new skill, profile, and claim plan to run at least one complete user journey:

    line 104
  16. Very specific:

    Run the owner's read-only Doctor and Spyglass Doctor.

    line 108
  17. Distinctive:

    Read each claimed side effect through a separate read-only view.

    line 110
  18. Distinctive:

    Keep failed attempts. Repair setup or the recipe, then repeat the affected claim against the final source.

    line 117
  19. Distinctive:

    If a claim or check changes, use an isolated negative control to show that the intended assertion fails.

    line 121
  20. Distinctive:

    If no fresh agent is available, read the skill as a new operator and remove steps that rely on chat history.

    line 125

“Common” is measured: kept by at least 6 in 10 of the 79 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Create a verification skill

Create a project-owned verify-<app> skill so the next agent can verify the app from a cold start. Use Spyglass for claims, checks, run state, and evidence. Keep the repository's own commands in charge of its installations, databases, and services.

1. Read the repository

Read repository instructions, README, manifests, task runners, existing verification skills, and feature maps. Read the relevant source for each user-facing surface. Answer these questions from the repository; ask the user only for facts the code cannot supply:

  • Surface: what does the user touch: web UI, API, CLI or TUI, desktop app, mobile app, library, or several of these?
  • Owner: which existing command prepares, checks, and cleans up an installation, database, service, or test identity? What resource ID proves the run owns it?
  • Drive: which public path can exercise the behavior? Use Spyglass's agent-browser session for UI journeys; keep existing browser tests, API, CLI, and library-facade journeys as named checks when they prove other behavior.
  • Observe: which independent reads, screenshots, responses, logs, or structured results prove each expected result?
  • Isolate: how can runs avoid a person's session, production, shared databases, and each other's state?

If the repository already has a verification skill, preserve its proven journeys and owner commands. Identify missing or stale coverage before editing. If the app cannot start or the owner check fails, record the concrete failure; do not describe the app as verified.

2. Add a Spyglass profile and claim plan

Read the installed CLI reference and schemas: spyglass --help, spyglass schema profile, and spyglass schema plan. In the Spyglass source, use schemas/profile.schema.json and schemas/plan.schema.json; use examples/demo.profile.json and examples/demo.plan.json for the current shape. Follow the installed CLI version when syntax differs.

If Spyglass is not installed, follow the Spyglass README's terminal setup, then check spyglass --version before creating the profile.

Run spyglass init when the repository has no profile. Review each proposed field and its provenance. Correct guesses from the owner's documentation and commands before starting a run. Configure only the capabilities needed by the selected journeys. Add the evidence directory to .gitignore when required.

Keep the repository owner in control of setup and cleanup. Use Spyglass's external launch mode to bind an owner-managed target. Add named checks for repeatable owner journeys and declare their structured result and artifacts. Use exec for a trusted owner command that has no named check. Spyglass records these operations; it does not replace the owner's lifecycle or authority.

Write a claim plan before driving. Each claim names the behavior, user entry point, preconditions, expected result, a relevant case, the named check, and required evidence. Include the affected consumer and a failure path. List excluded paths and their reasons. Mark guessed values as guesses until a live check confirms them.

Resolve credentials through secret references. Keep secret values out of the profile, plan, skill, feature map, helper, terminal arguments, and evidence.

3. Write the skill and feature map

Write SKILL.md with name: verify-<app> and a description that names the app, its main surface, and when to use the skill. Keep it available to agents. Give the next agent the exact owner setup, readiness, drive, independent-read, and cleanup steps. Use the same Spyglass CLI that a person can run in a terminal.

Create features/README.md and files for the top three to five user-facing features, selected from routes, commands, menus, or library operations. Include every changed feature and affected entry point. Keep each map user-focused and follow the examples in references/feature-map-example/README.md and create-note.md:

  1. Sub-features
  2. How to get to it (user POV)
  3. Driving it with <harness>
  4. Gotchas

For each user action, give the exact command or interaction and the observable result. Use stable accessible names, API paths, CLI prompts, or facade calls. Keep changing command syntax in Spyglass help and profile schemas; the local skill records only the repository's actual recipe.

Drive browser actions through Spyglass's owned browser session. Use existing Playwright, Cypress, or repository journeys as named checks when they cover specialist behavior. Do not build a second control CLI or write a throwaway drive script. Add a narrow reusable adapter only when a required operation is missing, then prove that adapter on the real target.

4. Prove one journey

Before handoff, use the new skill, profile, and claim plan to run at least one complete user journey:

  1. Prepare an isolated target with the repository owner's setup command.
  2. Run the owner's read-only Doctor and Spyglass Doctor.
  3. Drive the mapped user entry point and wait for asynchronous work to finish.
  4. Read each claimed side effect through a separate read-only view.
  5. Open and inspect actual screenshots with the current harness's image viewer; record the viewport, expected state, and observed state. If no operator can inspect the image, mark that visual claim blocked.
  6. Save the outcome and evidence, then clean up through the owner command and Spyglass. Confirm evidence remains after cleanup.

Keep failed attempts. Repair setup or the recipe, then repeat the affected claim against the final source. A screenshot file, successful command, request acceptance, or source review alone does not prove the journey.

If a claim or check changes, use an isolated negative control to show that the intended assertion fails. Restore the target and show that the same expectation passes. Record a precise limit when a safe control is unavailable.

If no fresh agent is available, read the skill as a new operator and remove steps that rely on chat history. Record that fresh-agent discovery remains untested.

5. Hand off

Report the skill and map paths, the proven claim and evidence, source identity, owner and Spyglass cleanup results, failed attempts, unverified paths, and limits. Name maintain-verification-skill as the upkeep pass when the app or its tools change. State whether a fresh agent followed the skill successfully.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.