agentleFS
Surething-ioSurething-io/cockpit/skills/spec/SKILL.md

spec

Skill44 starsChanged 8 months agoMITView on GitHub

Turn agreed requirements or a QA brief into an implementation-ready specification: behavior, contracts, boundaries, acceptance, and only the decisions that need the user. Use when asked to write or refine a SPEC, not to implement it.

168 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use backprop, ios-clean 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.
  • JuliusBrusseeSkill

    backprop

    Bug → spec protocol. When a bug is found or a test fails, trace the cause, decide whether a new §V invariant would catch recurrence, append to §B. This is the one non-obvious thing SDD does that plan-then-execute doesn't. Triggers on test failure, bug report, post-mortem, or explicit user ask.

    Also in 4 projects that use this

  • Saqlain-811Skill

    ios-clean

    Remove the DebugBridge SPM package and all #if DEBUG wiring from an iOS app. (gstack)

    Also in 5 projects that use this

  • Saqlain-811Skill

    ios-fix

    Autonomous iOS bug fixer. (gstack)

    Also in 5 projects that use this

  • bastosSkill

    ios-design-review

    Review iOS app UI/UX design quality using Apple's design principles and the Liquid Glass-era design language from WWDC25/26. Use when asked for a design review, UX critique, visual design assessment, or evaluation of an app's interface quality, design intent, alignment with the new Apple design system, or usability. Triggers: "design review", "UX review", "visual design critique", "review the UI", "does this design work", "evaluate this interface", "design feedback", "Liquid Glass review".

    Also in 5 projects that use this

  • ShreyRaviSkill

    plan-tune

    Self-tuning question sensitivity + developer psychographic for gstack (v1: observational). Review which AskUserQuestion prompts fire across gstack skills, set per-question preferences (never-ask / always-ask / ask-only-for-one-way), inspect the dual-track profile (what you declared vs what your behavior suggests), and enable/disable question tuning. Conversational interface — no CLI syntax required.

    Also in 7 projects that use this

  • timurgaleevSkill

    document-generate

    Generate missing docs for a feature, module or project as tutorials, how-tos, reference and explanation.

    Also in 6 projects that use this

  • +2 more from Saqlain-811

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 8 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 145 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read applicable repository instructions.

    line 13
  2. Distinctive:

    Check completeness and consistency, then stop.

    line 16
  3. Distinctive:

    Use fixed sections in this order.

    line 20
  4. Distinctive:

    Write for skimming: each rule has one home.

    line 43
  5. Distinctive:

    Use the QA-style condensed decision record.

    line 47
  6. Distinctive:

    Do not treat an answer as authorization to implement.

    line 76
  7. Distinctive:

    Before delivering, check that:

    line 82
  8. Distinctive:

    Every confirmed in-scope requirement has an observable acceptance outcome.

    line 84

“Common” is measured: kept by at least 6 in 10 of the 145 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

Specification mode: make agreed intent precise enough that development and testing do not have to invent product rules. Design only as far as the work needs. No implementation.

Input: a requirement, QA result, document path, or the agreement in the current conversation. QA is useful input, not a required skill dependency.

Workflow

  1. Extract confirmed requirements and decisions from the supplied sources and conversation. Keep current behavior, requested behavior, and proposed choices distinct; never present a recommendation as an agreement.
  2. Read applicable repository instructions. Inspect only the existing contracts, flows, and code needed to verify specific integration assumptions. Prefer named files and targeted searches; stop exploring once the specification has enough evidence. Cite material facts by source or path; label anything unverified.
  3. Resolve contradictions using explicit user corrections and confirmed decisions. If material sources still disagree, expose the conflict as a decision rather than silently choosing. Preserve earlier decisions unless new information actually invalidates them.
  4. Emit the SPEC using the structure below. Make useful progress even with open decisions; mark affected sections provisional. Ask only for choices that change product behavior, scope, compatibility, cost, or an important tradeoff. Choose routine implementation details using repository conventions and record consequential assumptions.
  5. Check completeness and consistency, then stop. Do not write application code, run migrations, deploy, send messages, or start implementation agents.

Output structure

Use fixed sections in this order; omit empty sections. Match the user's language and the repository's document conventions.

## Context          problem, goal, sources, verified current behavior; keep brief
## Scope            In / Out; affected actors, entry points, and surfaces
## Behavior         numbered rules (R1…); eligibility, ordering, states, transitions,
                    boundary conditions, and invariants relevant to this feature
## Interaction      entry → action → visible result; loading, empty, error, and
                    recovery states where relevant; reference supplied designs
## Contracts        affected interfaces and data; ownership, authorization,
                    validation, compatibility, and errors where relevant
## Design           proposed integration and responsibility boundaries; persistence,
                    concurrency, retries, and external dependencies only if needed
## Acceptance       scenarios (AC1…) linked to rules; precondition/action/expected
                    outcome, with verification method and external evidence needed
## Delivery         migration, configuration, rollout, and rollback requirements
                    only when this change needs them; distinguish release blockers
## Assumptions      unconfirmed defaults that materially affect what gets built
## Decisions        unresolved choices requiring the user, with recommendations

Scale to the work. A small UI change may need only Context, Scope, Interaction, and Acceptance. Do not manufacture APIs, data models, infrastructure, or deployment work to fill sections. When repository rules make a contract authoritative, reference it and specify the required changes rather than creating a competing source of truth.

Write for skimming: each rule has one home; use IDs to reference it elsewhere. Tables help compare states or outcomes. Include technical detail only when it closes an implementation gap or explains a tradeoff. Avoid line-by-line coding plans and speculative future extensibility.

Decisions

Use the QA-style condensed decision record. Number questions so the user can reply with letters alone. Do not re-ask answered questions or add a generic approval gate.

1. <question>

Drivers: <1–3 criteria>
A ★ <option>
    + <gain>
    - <cost>
B   <option>
    + <gain>
    - <cost>

★ A — <recommendation justified by the drivers>

Options must be mutually exclusive and cover the meaningful alternatives, at most three. End with “Reply 1A 2B, 'all as recommended', or correct anything above.” only when decisions remain.

Follow-up and saving

On answers or corrections, emit only the delta:

## Changed     <section/rule>: <old> → <new>; answered decisions as
               <chose X over Y> — <to achieve …>, accepting <cost>
## New         newly exposed assumptions or decisions, if any
## Open        still-open decision records; none → state that none remain

Do not treat an answer as authorization to implement. When asked for a complete or final SPEC, consolidate all accepted changes into one self-contained document, removing resolved questions and obsolete alternatives.

Default to delivering in conversation. If the user requests a saved artifact, use their path or the repository's established SPEC location; ask for a destination only if neither is available. Updating the requested SPEC is allowed; unrelated repository edits are not. Report the saved path. Do not create commits or branches unless requested.

Verification

Before delivering, check that:

  • Every confirmed in-scope requirement has an observable acceptance outcome; implementation suggestions are not reported as requirements.
  • Relevant failure and boundary cases have explicit outcomes, especially conflicting actions, permission boundaries, time cutoffs, and external failures.
  • Rules, interaction, contracts, design, and acceptance agree; no invented current behavior, template identifiers, credentials, or delivery guarantees.
  • External configuration or real-device validation remains explicitly pending until evidenced; writing a SPEC does not prove a feature works.
  • Material unknowns are visible, resolved decisions stay resolved, and no session-specific product details have leaked into this reusable workflow.

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.