Using Superpowers (Bootstrap)
The entry-point skill for the whole Superpowers system — mandates invoking a relevant skill whenever there's even a 1% chance it applies
Provides a useful session-start discovery and workflow entry point for coding agents using a skill library, reducing the chance of missing a relevant skill. Its boundary is session initialization and skill selection; the supplied material does not prove improved implementation quality. The mandatory pre-response rule has clear behavioral value but may over-process simple requests.
The prose rules are reasonably explicit and include tool mappings for Codex, Pi, and Antigravity. However, operation depends on host support for skill loading, subagents, and task tracking, and no general fallback is specified for other environments. There are no behavior tests for this skill, compatibility matrix, detailed conflict-resolution rules, or recovery procedure when invocation fails, so reproducibility evidence is limited.
The skill itself does not directly instruct command execution, credential access, or external service access, and it states that user instructions take precedence; no direct red-line issue is shown. Nevertheless, requiring invocation before every response or clarification can trigger unnecessary tools or subagents. It does not define pre-invocation permission checks, sensitive-data boundaries, approval for external effects, cancellation, or rollback. Telemetry information appears in the repository README, not as a complete safety control in this skill.
SKILL.md, its platform reference files, and the supplied repository tests support only a limited finding that tool mappings and documentation checks exist. The Antigravity test primarily checks strings and file presence rather than validating skill behavior. There is no independent evaluation, end-to-end invocation evidence, failure-case coverage, or traceable support for performance or quality claims; README methodology claims remain repository assertions rather than independent facts.
The name, description, core rule, and some platform adaptations are discoverable, and normal use requires little user input. However, the material lacks a clear input/output contract, post-install verification steps, troubleshooting for failed invocation, and an explicit exit path. The references partially explain harness differences but do not cover all compatibility limits, costs, or skill-conflict cases.
The repository supplies an MIT license, package.json version 6.1.1, contribution guidance, update information, and an issue channel, providing some maintenance signals. The selected skill has no own version, changelog, owner, or response commitment, and the README statement that updates are 'often automatic' is not a concrete update path. The publisher is unverified in the supplied registry context, so it cannot add governance or maintenance credit.
- Do not treat 'invoke before any response' as blanket authorization; confirm host capabilities, permissions, and user intent before tools, subagents, file changes, or external actions.
- Invocation may fail when the current harness lacks the documented skill-loading, task-tracking, or subagent capabilities; provide a manual or sequential fallback.
- Do not treat repository README claims about methodology effectiveness, testing, or platform coverage as independently verified evidence for this selected skill.
What it does & when to use it
The entry point for the entire Superpowers system. It doesn't do concrete work itself — its job is to establish, at the start of every session (and again after every context compaction), a hard rule: if a skill might apply, even at 1% confidence, invoke it first — including before clarifying questions or exploring the codebase, actions that feel like they haven't "really started" yet. The doc is short (62 lines); its core is a rationalization table that rebuts excuses like "this is just a simple question" or "I already know what that skill does" one by one.
Sets the invocation order: when multiple skills apply, process skills (brainstorming, systematic-debugging) come before implementation skills. Gives two concrete trigger examples ("let's build X" → brainstorming first; "fix this bug" → systematic-debugging first). Lists a table of thoughts that signal you're rationalizing your way out of a check. States that user instructions (CLAUDE.md/AGENTS.md/direct requests) outrank skills, and skills outrank default behavior — a skill or instruction can only be skipped on an explicit user request. Explicitly exempts subagents dispatched to execute a specific task — they don't need to run this "check for a relevant skill" layer at all.
- The mandatory first check at the start of any new session — before any clarifying question or codebase exploration, determine whether a relevant skill exists
- Explicitly skipped when dispatched as a subagent executing a specific task (the SUBAGENT-STOP marker), so subagents don't recursively re-trigger "should I use Superpowers" checks
- Re-established after a session undergoes context compaction, so the rule isn't forgotten once the conversation gets summarized
Pros & cons
- States the "even 1% chance, still mandatory" rule in absolute terms, closing off the fuzzy self-exemptions of "this doesn't count" or "I already know this skill"
- Clearly distinguishes the main-session role from the subagent-dispatched-for-a-task role, preventing recursive unnecessary checks when subagents dispatch further subagents
- Explicitly states that direct user instructions outrank the skill rule, leaving a reasonable escape hatch for human override rather than being fully non-negotiable
- Only reliably self-enforcing on harnesses that actually read and follow SKILL.md instructions (Claude Code, Codex CLI, etc.) — in a plain chat interface it's just a document a human has to apply by hand.
- This kind of "mandatory self-check" rule has no technical enforcement mechanism — it depends entirely on the model actually following it, and is toothless the moment the model selectively ignores it
- Running a "is there a relevant skill" check before every action adds real decision overhead on simple tasks — it isn't free
How to install
Installed as part of the obra/superpowers plugin — individual skills can't be installed standalone:
- Claude Code: official marketplace
/plugin install superpowers@claude-plugins-official; or register the community marketplace with/plugin marketplace add obra/superpowers-marketplacethen/plugin install superpowers@superpowers-marketplace - Also supports Antigravity, Codex App/CLI, Cursor, Factory Droid, GitHub Copilot CLI, Kimi Code, OpenCode, and Pi — see the repo README for the exact command per harness
Once installed, this skill activates alongside the other 13 Superpowers skills, auto-triggered by using-superpowers based on context — there's no separate invocation step.
How to use
No manual trigger needed — it's the plugin's bootstrap entry point, meant to re-take effect at the start of every new session or after every compaction so all following behavior checks for a relevant skill before acting.
Compared to similar skills
Compared to having no bootstrap skill and relying entirely on the model to decide on its own whether to check for relevant documentation, this skill turns "check for a skill" from optional into a mandatory first step. Compared to a one-line system-prompt reminder like "prefer available tools," this gives concrete trigger examples, a rationalization table, and an explicit role distinction — considerably higher rule density and actionability.