quoin

Finalizes a completed task: ensures all changes are committed, pushes branch to remote, prompts for lessons learned, aggregates task cost across all sessions, and marks the task as complete. Requires /review to have been run first. Does NOT create a PR — that's a separate explicit action. Use this skill for: /end_of_task, 'finalize this', 'we're done', 'ship it', 'task complete', 'wrap up this task'. This is the explicit user acceptance of completed, reviewed work — the last step before moving on.

Implementation agent that executes tasks from a plan. Uses Sonnet for efficient, high-quality code generation. Use this skill for: /implement, implementing a plan, writing code from a plan, executing implementation tasks, 'implement task N from the plan', 'start coding', 'build this based on the plan'. Triggers whenever the user wants to turn a plan into actual code changes.

Create a pull request: optional version bump, push branch if not already pushed, create PR via gh, wait for merge, switch to merge target. Use this skill for: /pr, 'create a PR', 'open a pull request', 'submit for review'. Triggers when the user wants to create a PR from a feature branch.

Safely undo an implementation phase or revert specific tasks, using the plan to map commits to tasks. Use this skill for: /rollback, 'undo the implementation', 'revert the last changes', 'go back to before implement', 'undo task 3', 'reset to pre-implementation'. Reads the plan to understand which commits belong to which tasks, shows what would be reverted, and requires explicit confirmation before acting.

Returns a live cost summary showing today's cost, project lifetime cost, and per-open-task breakdown. Use for: /cost_snapshot, 'how much have I spent', 'cost report', 'show costs', 'what's the project cost', 'how much has this task cost'. Read-only — no file artifacts produced.

Consolidates all of today's work across all sessions into a daily cache for next-day resumption. Works in any session — fresh or active. Use this skill for: /end_of_day, 'wrapping up', 'done for the day', 'save my progress', 'end of day', 'EOD'. Captures what was worked on, what's unfinished, blockers, decisions made, and recent git activity. The daily cache feeds into /start_of_day for seamless resumption.

Restores context from the daily cache and unfinished sessions so you can resume where you left off. Use this skill for: /start_of_day, 'what was I working on', 'resume', 'pick up where I left off', 'morning standup', 'SOD', 'start of day'. Reads the latest daily cache, checks git state, and presents a clear picture of what to do next.

Renders the quoin workflow pipeline graph with the active phase marked. Use for: /status, 'where am I', 'what phase am I in', 'show workflow progress', 'pipeline graph', 'workflow status'. Read-only — no file artifacts produced.

Routes a natural-language prompt to the right workflow skill. Reads the prompt, inspects workflow state, classifies intent against the full skill catalog, proposes the best-fit skill with rationale, then tells the user which command to type. Use for: /triage, 'what should I run', 'which skill fits this', 'route this', 'pick the right command for me', 'I'm not sure what to do next'. Never invokes skills directly — always propose-only.

Aggregates the week's meaningful work into a structured review. Use this skill for: /weekly_review, 'weekly summary', 'what did I do this week', 'week recap', 'friday review', 'weekly standup', 'weekly report'. Reads daily caches, session files, git history, and lessons learned to produce a comprehensive but concise picture of the week's progress, decisions, and outcomes.

Consolidates all of today's work across all sessions into a daily cache for next-day resumption. Works in any session — fresh or active. Use this skill for: /end_of_day, 'wrapping up', 'done for the day', 'save my progress', 'end of day', 'EOD'. Captures what was worked on, what's unfinished, blockers, decisions made, and recent git activity. The daily cache feeds into /start_of_day for seamless resumption.

Scans all repositories in the project folder and saves a comprehensive inventory, architecture overview, and dependency map to memory. Use this skill for: /discover, 'scan my repos', 'map the codebase', 'what repos do I have', 'index the project', 'save the architecture', 'learn my codebase'. Run this when you first set up the workflow in a new project folder, or when repos have changed significantly. The output feeds into /architect, /plan, /critic, and /review so they have baseline context.

Initializes the development workflow in a project folder. Creates the .workflow_artifacts/ structure, runs /discover to scan the codebase, and generates a quickstart guide. Requires install.sh to have been run first (installs skills to __QUOIN_HOME__/skills/ and workflow rules to __QUOIN_HOME__/CLAUDE.md). Use this skill for: /init_workflow, 'initialize workflow', 'set up dev workflow', 'install workflow', 'bootstrap workflow'. Run this once per project.

General-purpose state-saving — save session-restore state mid-session before context exhaustion, between tasks, between sessions, or before starting new heavy work. Writes a pending-restore sentinel so a fresh session can resume exactly where you left off. Also surfaces pending-restore state on --restore. Auto-runs /cleanup on save (trash-moves stale sentinels/old checkpoints) unless --no-cleanup. Use for: /checkpoint, 'save my place', 'checkpoint', 'save session', '/checkpoint --restore', 'restore checkpoint', 'resume from checkpoint'. Does NOT roll up dailies, does NOT touch lessons-learned.md or forgotten/.

Triages tasks by size (Small/Medium/Large) and orchestrates the appropriate planning path. Small tasks get a single-pass /plan (no critic loop). Medium tasks run the plan→critic→revise cycle with Sonnet revision. Large tasks (or 'strict:' prefix) run all-Opus with up to 5 rounds. Use this skill for: /thorough_plan, 'plan this', 'plan this thoroughly', 'detailed plan with review', 'plan and critique', 'full planning cycle'. Supports size tags (small:/medium:/large:), strict: prefix, and max_rounds: N override. Always the entry point for planned work — routes automatically based on task size.

Memory consolidation skill — scans daily insights and session files, promotes patterns to lessons-learned.md, soft-forgets stale entries to forgotten/. Auto-invoked by /end_of_day as its final step. Standalone: /sleep [--dry-run] [--quiet-forget] [--escalate] [--skip-sleep] [--restore <pattern>] [--purge --older-than 90d].

Trash-move stale workflow sentinels and old checkpoints into recoverable archive. Keeps the freshest/current session's sentinels; trashes others older than QUOIN_CLEANUP_SENTINEL_WINDOW (default 1d). Trashes checkpoints older than QUOIN_CLEANUP_CKPT_WINDOW (default 30d). Auto-fires as the first sub-block of Step 1.5 in /checkpoint (default-on, --no-cleanup opt-out). Standalone: /cleanup [--dry-run]. Recovery via manual mv from .workflow_artifacts/memory/trash/<date>/ — NOT /sleep --restore.

Resume context from a prior session. Reads recent-sessions.md to identify session IDs, lets you pick a session to revive, then extracts the last checkpoint summary and recent messages from the session JSONL. Use for: /continue_work, 'resume prior session', 'revive old session', 'where was I', 'switch to last session'.

Append-only queue for future work items. Use for: /next-steps, /next-steps add <text>, /next-steps list, /next-steps list --all, /next-steps done N. Manages a ## Queue section in next-steps.md at the project root.

General-purpose state-saving — save session-restore state mid-session before context exhaustion, between tasks, between sessions, or before starting new heavy work. Writes a pending-restore sentinel so a fresh session can resume exactly where you left off. Also surfaces pending-restore state on --restore. Auto-runs /cleanup on save (trash-moves stale sentinels/old checkpoints) unless --no-cleanup. Use for: /checkpoint, 'save my place', 'checkpoint', 'save session', '/checkpoint --restore', 'restore checkpoint', 'resume from checkpoint'. Does NOT roll up dailies, does NOT touch lessons-learned.md or forgotten/.

Trash-move stale workflow sentinels and old checkpoints into recoverable archive. Keeps the freshest/current session's sentinels; trashes others older than QUOIN_CLEANUP_SENTINEL_WINDOW (default 1d). Trashes checkpoints older than QUOIN_CLEANUP_CKPT_WINDOW (default 30d). Auto-fires as the first sub-block of Step 1.5 in /checkpoint (default-on, --no-cleanup opt-out). Standalone: /cleanup [--dry-run]. Recovery via manual mv from .workflow_artifacts/memory/trash/<date>/ — NOT /sleep --restore.

Resume context from a prior session. Reads recent-sessions.md to identify session IDs, lets you pick a session to revive, then extracts the last checkpoint summary and recent messages from the session JSONL. Use for: /continue_work, 'resume prior session', 'revive old session', 'where was I', 'switch to last session'.

Append-only queue for future work items. Use for: /next-steps, /next-steps add <text>, /next-steps list, /next-steps list --all, /next-steps done N. Manages a ## Queue section in next-steps.md at the project root.

Memory consolidation skill — scans daily insights and session files, promotes patterns to lessons-learned.md, soft-forgets stale entries to forgotten/. Auto-invoked by /end_of_day as its final step. Standalone: /sleep [--dry-run] [--quiet-forget] [--escalate] [--skip-sleep] [--restore <pattern>] [--purge --older-than 90d].

Deep architectural analysis and planning using the strongest available model (Opus), with a scan/synthesize split for efficiency. Spawns Sonnet subagents in parallel to read repos, then synthesizes findings on Opus. Use this skill whenever the user needs to explore a complex system, understand how multiple repositories interact, design a new architecture, decompose a large problem into implementable stages, or answer hard cross-cutting questions about a codebase. Triggers on: /architect, architecture design, system design, technical exploration, cross-repo analysis, complex technical questions, 'how should we build this', 'what's the best approach for', deep code exploration, multi-service design. Even if the user just says 'I need to think through X' where X is technical — use this skill.

Create a detailed, implementation-ready plan for a development task using the strongest model (Opus). Use this skill for: /plan, 'plan this', 'create a plan', 'break this down into tasks', 'how should we implement', implementation planning. Produces a concrete plan with task decomposition, integration analysis, risk assessment, and testing strategy. Can be used standalone or as part of /thorough_plan orchestration.

Senior technical critic that reviews implementation plans for gaps, risks, and integration issues using the strongest model (Opus). Use this skill for: /critic, 'critique this plan', 'review the plan', 'find issues with this plan', 'what's wrong with this approach'. The critic reads both the plan AND the actual codebase to catch mismatches. Can be used standalone or as part of /thorough_plan orchestration.

Deep code review using the strongest model (Opus) to verify implementation matches the plan and is production-ready. Use this skill for: /review, code review, review implementation, check if code matches plan, verify implementation, 'does this look right', 'review my changes', 'check the implementation', post-implementation review. Triggers whenever the user wants to validate that implemented code is correct, complete, and safe.

Revises an implementation plan based on critic feedback, addressing all critical and major issues using the strongest model (Opus). Use this skill for: /revise, 'fix the plan', 'address the critic's comments', 'update the plan based on feedback'. Reads the critic response, updates current-plan.md, and documents what changed. Can be used standalone or as part of /thorough_plan orchestration.

Captures a quick insight, pattern, or discovery to the daily insights scratchpad without interrupting the current task. Use for: /capture_insight, 'note this', 'remember this pattern', 'log this decision', 'save this as a lesson', 'note that', 'remember that'. Appends a structured entry to .workflow_artifacts/memory/daily/insights-<date>.md.

Expands compressed (terse) workflow artifacts back to English for human reading. Use for: /expand <path>, 'show me the English version of', 'expand this file', 'what does this terse file say'. Dispatches: Class B summary detection (reads ## For human block at top of v3 artifacts), no-op display (Tier 1 English files), LLM re-expansion (Tier 3 ephemeral files — lossy, banner-flagged). Never used as a contract approval path.

Automated quality gate that runs checks and requires explicit human approval before the workflow can proceed to the next phase. Use this skill for: /gate, 'check before proceeding', 'run the gate', 'verify before next step'. Runs lint, typecheck, tests, and presents a summary with go/no-go decision to the user. No phase transition happens without the user's explicit approval. This is a blocking checkpoint — the workflow STOPS here until the user says go.

Fast variant of /revise using Sonnet for cost-efficient plan revision. Content-identical to /revise but runs on Sonnet instead of Opus. Used by /thorough_plan in default (non-strict) mode for rounds 2-3. Not intended for direct user invocation — use /revise for standalone revision.

Deep architectural analysis and planning using the strongest available model (Opus), with a scan/synthesize split for efficiency. Spawns Sonnet subagents in parallel to read repos, then synthesizes findings on Opus. Use this skill whenever the user needs to explore a complex system, understand how multiple repositories interact, design a new architecture, decompose a large problem into implementable stages, or answer hard cross-cutting questions about a codebase. Triggers on: /architect, architecture design, system design, technical exploration, cross-repo analysis, complex technical questions, 'how should we build this', 'what's the best approach for', deep code exploration, multi-service design. Even if the user just says 'I need to think through X' where X is technical — use this skill.

Senior technical critic that reviews implementation plans for gaps, risks, and integration issues using the strongest model (Opus). Use this skill for: /critic, 'critique this plan', 'review the plan', 'find issues with this plan', 'what's wrong with this approach'. The critic reads both the plan AND the actual codebase to catch mismatches. Can be used standalone or as part of /thorough_plan orchestration.

Create a detailed, implementation-ready plan for a development task using the strongest model (Opus). Use this skill for: /plan, 'plan this', 'create a plan', 'break this down into tasks', 'how should we implement', implementation planning. Produces a concrete plan with task decomposition, integration analysis, risk assessment, and testing strategy. Can be used standalone or as part of /thorough_plan orchestration.

Deep code review using the strongest model (Opus) to verify implementation matches the plan and is production-ready. Use this skill for: /review, code review, review implementation, check if code matches plan, verify implementation, 'does this look right', 'review my changes', 'check the implementation', post-implementation review. Triggers whenever the user wants to validate that implemented code is correct, complete, and safe.

Fast variant of /revise using Sonnet for cost-efficient plan revision. Content-identical to /revise but runs on Sonnet instead of Opus. Used by /thorough_plan in default (non-strict) mode for rounds 2-3. Not intended for direct user invocation — use /revise for standalone revision.

Revises an implementation plan based on critic feedback, addressing all critical and major issues using the strongest model (Opus). Use this skill for: /revise, 'fix the plan', 'address the critic's comments', 'update the plan based on feedback'. Reads the critic response, updates current-plan.md, and documents what changed. Can be used standalone or as part of /thorough_plan orchestration.

Renders the quoin workflow pipeline graph with the active phase marked. Use for: /status, 'where am I', 'what phase am I in', 'show workflow progress', 'pipeline graph', 'workflow status'. Read-only — no file artifacts produced.