AgentStack
Browse Sign in
Browse Why AgentStack Sell Docs
Sign in
SKILL verified Apache-2.0 Self-run

Ss Feature Specification

skill-jonisavo-supersigil-ss-feature-specification · by jonisavo

Default skill for authoring or repairing Supersigil specification documents: requirements, design docs, tasks, and ADRs. Use when turning an agreed feature idea or imported material into verifiable docs, refining acceptance criteria, or bringing specs to an implementation-ready state. Works alongside brainstorming and planning skills.

— No reviews yet
0 installs
32 views
0.0% view→install

Install

$ agentstack add skill-jonisavo-supersigil-ss-feature-specification

✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • ✓ Prompt-injection patterns
  • ✓ Secret / credential exfiltration
  • ✓ Dangerous shell & filesystem operations
  • ✓ Untrusted network calls
  • ✓ Known-malicious package signatures

What it can access

  • ✓ Network access No
  • ✓ Filesystem access No
  • ✓ Shell / process execution No
  • ✓ Environment & secrets No
  • ✓ Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README, it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-jonisavo-supersigil-ss-feature-specification)

Reliability & compatibility

✓ Security review passed
0 installs to date
— no reviews yet
● 3mo ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Ss Feature Specification? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Feature Specification

Use this skill to create or repair Supersigil spec documents for one bounded feature or prefix. Augment the user's existing brainstorming or planning workflow; do not invent product direction they have not asked for.

Current Contract

Use the current CLI as the source of truth:

supersigil import --from kiro ...
supersigil new  
supersigil schema [--format json]
supersigil verify
supersigil ls [--format json]
supersigil context  [--format json]
supersigil plan [] [--format json]
supersigil verify [--format terminal|json|markdown]
supersigil status [] [--format json]
supersigil affected --since  [--format json]
supersigil graph [--format mermaid|dot]

Prefer supersigil new over hand-writing boilerplate when starting from scratch. Prefer supersigil import --from kiro when the user already has .kiro/specs material. Use supersigil schema to discover the current component and document type surface before inventing structure. Use [references/templates.md](references/templates.md) only as a fallback example set when new is too sparse or imported prose needs normalization. Use [references/multi-project.md](references/multi-project.md) when working in a monorepo with [projects.*] entries.

Write string literal attributes only. Even if schema examples still show JSX expression syntax like refs={["..."]}, author docs as refs="a, b" and paths="x, y". The parser and lint rules reject expression attributes.

Workflow

  1. Bound the scope to one feature, subsystem, or import batch.

Do not try to normalize the whole repo at once.

  1. Inspect the current state before editing.

Run supersigil schema --format json to get the current component and document type definitions. Run supersigil ls --format json to see what already exists for the feature. Run supersigil context and supersigil status on the main requirement or design doc when one already exists. Run supersigil plan to see uncovered criteria and pending tasks.

  1. Prefer existing material over blank documents.

If .kiro/specs exists and the user wants Supersigil docs, import first, then refine the generated MDX. If Supersigil docs already exist, inspect them before rewriting. If no Supersigil docs exist yet, scaffold with supersigil new.

  1. Keep every authored document in status: draft.

Use later statuses only when the document actually earned them. Do not promote requirement or design docs based only on lint. Treat draft as the safe working state until verify is clean and the user has reviewed the result.

  1. Run the requirements phase first.

Author or refine the requirement doc before design or tasks. Reuse imported or existing prose when it is already good enough. Start from supersigil new scaffolds, then expand them with the patterns in [references/templates.md](references/templates.md) when they need richer structure.

  1. Pause for a requirement feedback round.

Show the user the drafted requirement criteria. Ask for feedback before moving on to design. Refine the requirement-phase docs based on that feedback instead of pushing ahead.

  1. Run the design phase second.

Author the design doc only after the requirement feedback round has been incorporated. Keep the design tied to the reviewed requirement shape.

  1. Record design decisions where they belong.

When a feature involves only a few decisions, embed `, , and components directly in the design doc. For large, scoped decisions with non-obvious tradeoffs, rejected alternatives worth recording, or choices that future contributors will question, suggest a dedicated ADR (supersigil new adr /). Link decisions to requirement criteria with . Use standalone="..." on decisions that have no corresponding requirement. ADR statuses are draft → review → accepted → superseded`.

  1. Pause for a design feedback round.

Show the user the design direction, boundaries, and major implementation choices. Include any ADRs authored in this round. Refine the design doc and ADRs before creating tasks.

  1. Run the tasks phase last.

Create dependency-ordered tasks from the current requirement and design graph. Keep tasks actionable and criterion-linked.

  1. Offer a tasks feedback round.

Let the user adjust task granularity, ordering, or scope before handoff to implementation. Refine the tasks doc if the user wants another pass.

  1. Run supersigil verify after every write.

Treat lint cleanliness as the minimum quality bar. If lint fails, fix that before doing more structural work.

  1. Rebuild the graph after structural edits.

Run supersigil ls, supersigil context, supersigil plan, or supersigil graph after adding or changing refs. If these commands fail after a change, assume you broke cross-document refs or task dependencies and fix that immediately.

  1. Run supersigil verify before claiming the feature spec is ready.

Fix uncovered criteria, missing test mappings, stale tracked files, and status inconsistencies that matter for the scoped feature. If the user is still drafting and wants to defer a finding, keep the relevant doc at status: draft and state the remaining gap clearly.

  1. Use supersigil status to decide handoff readiness.

Project the next step from the actual state: more spec work, human review, or implementation. Use supersigil affected --since when the user wants to review which docs source changes may have invalidated.

  1. End with a concrete handoff.

Summarize which docs are lint-clean, which docs are verify-clean, which review rounds were completed, which IDs or prefixes to inspect with context, plan, and status, and what remains open before execution.

Authoring Rules

  • Use stable IDs and match the current repo convention unless the user has a stronger local convention.

The current built-in scaffolds shorten only requirement to req, so common primary IDs are {feature}/req, {feature}/design, {feature}/tasks, and {feature}/adr. Keep criterion and task IDs stable once other docs reference them. Numeric IDs such as req-1-1 and task-2-3 are acceptable.

  • Keep relationship direction concrete to abstract.

Requirement docs own ` entries. Design docs point to requirement docs with . Tasks point to criteria with the implements attribute on each . ADR docs use , , and components and can link to criteria with `.

  • Put acceptance criteria only in requirement docs.
  • Make tasks dependency-ordered with depends and keep them actionable.
  • Use `` for document-level ordering only.
  • Write list attributes as comma-separated strings, not JSX expressions.

Use refs="doc#a, doc#b" and paths="src/**/*.rs, tests/**/*.rs". Never write refs={...} or paths={...}.

  • Never emit empty placeholders like refs="", paths="", or tag="".

Omit the component until a real value exists.

  • Use `` with the strategies the verification engine understands today.

Prefer strategy="tag" when tests carry supersigil: ... comments. Use strategy="file-glob" when concrete test file globs are known.

  • Use `` only when concrete source paths are already known.

Failure Modes

  • Do not invent components or attributes. Use supersigil schema --format json as the source of truth, and fall back to [templates.md](references/templates.md) only if the command is unavailable.
  • Do not trust schema example syntax over the parser. Expression attributes are illustrative noise right now; lint is authoritative.
  • Do not mark work complete because imported docs exist. Imported docs are starting material, not proof of correctness.
  • Do not leave broken refs in place for later. context and plan depend on a loadable graph.
  • Do not spread one feature across many unrelated IDs or folders without a clear prefix strategy.
  • Do not promote statuses optimistically just because lint is clean. Use verify, status, and human review to justify status changes.
  • Do not skip the user review checkpoints between requirements, design, and tasks unless the user explicitly asks for a faster pass.

Handoff

When requirements, design, and tasks have each had their review pass, and the scoped docs are lint-clean and verify-clean enough for honest execution, suggest ss-feature-development if that skill exists. Otherwise hand off by pointing the user to:

  • supersigil context for relationship review
  • supersigil plan for outstanding work
  • supersigil status for readiness review
  • supersigil verify for the final spec health check
  • the edited docs for human review

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet, be the first.

Versions

  • v0.1.0 Imported from the upstream source.