Superpowers Mode

Enable or disable a strict engineering workflow for coding tasks enforcing goal clarification, specs, planning, test-driven small steps, and verification.

Security Scan

This instruction-only skill is internally consistent with its stated purpose: it enforces a coding workflow, stores a small mode file in agent memory, and does not request credentials or install software.

Purpose & Capability

Name, description, and instructions align: a workflow helper for coding tasks that toggles behavior and uses reference templates. There are no unrelated env vars, binaries, or installs. Note: the skill has no public homepage or source URL (owner is an ID only), which reduces transparency but does not create a technical mismatch.

Instruction Scope

SKILL.md stays within scope: it reads/writes a single state file (memory/superpowers-mode.md), produces specs/plans, and restricts itself to coding/build/debug tasks. It does not request other files, credentials, or external endpoints. Note: it persists mode state in agent memory, which is expected behavior for a toggle.

Install Mechanism

No install spec and no code files — lowest-risk model (instruction-only). Nothing is downloaded or written to disk beyond the declared state file in agent memory.

Credentials

No environment variables, credentials, or config paths are requested. The declared requirements are minimal and appropriate for the stated functionality.

Persistence & Privilege

The skill writes a small state file to memory (enabled:true/false) and confirms changes; it does not require always:true and does not modify other skills or system settings. Autonomous model invocation is allowed by default (disable-model-invocation is false), which is normal for skills and not by itself a concern.

Assessment

This skill appears coherent and low-risk: it only enforces a development workflow and stores a small mode file in agent memory. Before installing, consider: (1) the skill has no public homepage or source URL — if provenance matters to you, prefer skills with a visible maintainer; (2) the state file (memory/superpowers-mode.md) is persistent within the agent environment — if your agent shares memory across skills or sessions, check how that storage is protected; (3) the skill requests no credentials and performs no network or shell actions, so it cannot exfiltrate secrets on its own. If you’re comfortable with the unknown owner and agent memory usage, this skill is reasonable to enable.

Superpowers Mode (On-demand)

Use this only when the user explicitly asks to enable/disable this mode, or when mode is already enabled for coding tasks.

State file

Track mode in:

memory/superpowers-mode.md

Format:

enabled: true|false
updatedAt: <ISO>
notes: <optional>

Commands

• Enable phrases: включи superpowers, enable superpowers, superpowers on
  • Write enabled: true to the state file.
  • Confirm in 1 short message.
• Disable phrases: выключи superpowers, disable superpowers, superpowers off
  • Write enabled: false.
  • Confirm in 1 short message.
• Status phrases: статус superpowers, superpowers status
  • Read state and report enabled/disabled.

Workflow when enabled (coding tasks only)

For coding/build/debug requests, follow this order:

1. Clarify objective and constraints quickly.
2. Produce a short spec (chunked, easy to review).
3. Produce an implementation plan with small tasks.
4. (Optional, 30 sec) Run a mini risk review:
   • How can this fail in production?
   • What is the weakest dependency/state assumption?
   • What signal will show regression + how to rollback fast?
5. Execute task-by-task (prefer test-first for risky changes).
6. Verify against acceptance criteria, then summarize outcome + next step.

Use templates from references/ when useful.

Red flags (quick self-check)

If you notice these thoughts, slow down and apply the workflow:

• "Сейчас быстро вкачу без плана" for non-trivial changes.
• "И так понятно, тесты потом" on risky edits.
• "Откат не нужен" before touching config/auth/cron/system files.
• "Похоже работает" without explicit verification.

Guardrails

• Do not force this workflow for non-coding chat.
• If user asks for speed (quick, без плана, just do it), skip to minimal plan and execute.
• Keep updates concise; avoid process spam.