ワンクリックで
feature
// Add a feature to an existing project with a full sprint. Skips /think diagnostic, goes straight to planning. Use when the user knows what they want and the project already exists. Triggers on /feature.
// Add a feature to an existing project with a full sprint. Skips /think diagnostic, goes straight to planning. Use when the user knows what they want and the project already exists. Triggers on /feature.
Use when working near production, sensitive systems, or destructive operations. Activates on-demand safety hooks that block dangerous commands. Supports modes — careful (warn), freeze (guided, keeps writes within scope), unfreeze (remove restrictions). Triggers on /guard, /careful, /freeze, /unfreeze.
Orchestrate parallel agent sessions through a sprint. Coordinates task claiming, dependency resolution, and artifact handoff between independent agents. Triggers on /conductor, /sprint, /parallel.
Use to verify that code works correctly — browser-based testing with Playwright, native app testing with computer use, CLI testing, API testing, or root-cause debugging. Supports --quick, --standard, --thorough modes. Triggers on /qa.
Use when starting non-trivial work (touching 3+ files, new features, refactors, bug investigations). Produces a scoped, actionable implementation plan before any code is written. Triggers on /nano.
Use after writing code to get a thorough code review. Runs two passes — structural correctness then adversarial edge-case hunting. Scales depth by diff size. Supports --quick, --standard, --thorough modes. Triggers on /review.
Use before shipping to production. Performs OWASP Top 10 audit and STRIDE threat modeling against the codebase. Supports --quick, --standard, --thorough modes. Also use when the user asks to check security, audit code, or review for vulnerabilities. Triggers on /security.
| name | feature |
| description | Add a feature to an existing project with a full sprint. Skips /think diagnostic, goes straight to planning. Use when the user knows what they want and the project already exists. Triggers on /feature. |
| concurrency | read |
| depends_on | [] |
| summary | Fast sprint for incremental features. Reads existing artifacts, plans, builds, reviews, audits, ships. |
| estimated_tokens | 200 |
| hooks | {"PreToolUse":[{"matcher":"Bash","command":"./feature/bin/enforce-sprint.sh"}]} |
Fast path for adding a feature to an existing project. Skips the /think diagnostic and runs the full sprint via skill invocations.
/feature Add import from JSON/CSV to restore backups
Defensive telemetry init. No-op if telemetry is disabled via NANOSTACK_NO_TELEMETRY=1, ~/.nanostack/.telemetry-disabled, or if the helpers are removed.
_P="$HOME/.claude/skills/nanostack/bin/lib/skill-preamble.sh"
[ -f "$_P" ] && . "$_P" feature
unset _P
Before anything else, ensure the project is configured. Run this once (skips if already done):
[ -f .claude/settings.json ] || ~/.claude/skills/nanostack/bin/init-project.sh
Initialize the sprint session with autopilot and explicit plan auto-approval. --autopilot already implies --plan-approval auto per the session contract, but passing it explicitly makes the intent obvious to anyone reading session.json.
~/.claude/skills/nanostack/bin/session.sh init feature --autopilot --plan-approval auto
Manual feature work should use /think + /nano instead. /feature itself does not accept a manual mode flag.
Then run session.sh phase-start plan. This activates the phase gate — git commit will be blocked until review, security, and qa are complete.
You are an autonomous orchestrator. You run the entire sprint without stopping between phases. Do NOT wait for user input between steps. Do NOT ask "should I continue?" or "ready for review?". Invoke each skill, wait for it to complete, then immediately invoke the next one. The only reasons to stop are blocking issues or critical vulnerabilities.
Auto-approval contract for sub-skills. The session records plan_approval=auto and autopilot=true, so /nano, /review, /security, /qa, and /ship read those fields and behave accordingly: present briefly, do not pause for approval. If any of them asks "ready to proceed?", treat it as a regression in that skill, not a signal to stop.
Resolve existing artifacts and solutions in one call:
~/.claude/skills/nanostack/bin/resolve.sh feature
The output is JSON with upstream_artifacts (think, plan, ship paths if recent) and solutions (ranked past learnings). Read the checkpoint summaries. If no artifacts exist, read the codebase directly.
Use Skill tool: skill="nano"
Wait for /nano to complete. It saves its own artifact. Then immediately build.
Build the feature. Do not ask for approval. The plan was the contract.
These three phases are independent. They all read the build output but don't depend on each other. Launch all three using the Agent tool in a single message with three parallel tool calls:
Agent: subagent_type="general-purpose", prompt="Run /review on this project. Use Skill tool: skill='review'"
Agent: subagent_type="general-purpose", prompt="Run /security on this project. Use Skill tool: skill='security'"
Agent: subagent_type="general-purpose", prompt="Run /qa on this project. Use Skill tool: skill='qa'"
If parallel agents are not available, fall back to sequential:
Use Skill tool: skill="review"
Use Skill tool: skill="security"
Use Skill tool: skill="qa"
If any phase finds blocking issues or critical vulnerabilities: fix them, then re-run that phase only.
Feature: review + security + qa complete. Running /ship...
Use Skill tool: skill="ship"
/ship commits, creates PR if remote exists, generates sprint journal, runs /compound, and shows the result with next feature suggestions.
Before returning control:
_F="$HOME/.claude/skills/nanostack/bin/lib/skill-finalize.sh"
[ -f "$_F" ] && . "$_F" feature success
unset _F
Pass abort or error instead of success if the feature flow did not complete normally.
Feature: review complete. Running /security...