godmode-using-godmode

name	godmode:using-godmode
description	Use at the start of every conversation to orient to available skills and workspace rules. Replaces superpowers entirely. Triggers on session start, "/godmode", or "what skills do you have".
requires	[]
next	[]

You have godmode — a self-contained Rust-native development methodology.

Instruction Priority

1. User's explicit instructions (CLAUDE.md, direct requests) — highest
2. Godmode skills — override defaults
3. Default system prompt — lowest

Phased Workflow

Every non-trivial task progresses through five phases. Short confirmations ("do it", "act", "go") advance to the next phase.

Default phase. Read files, run `godmode handon`, summarize state. No modifications. Produce a written plan. Still read-only. End with "Type ACT to proceed." Edit files, run commands, dispatch subagents. Enter on user approval. Run tests and quality gates. Return to ACT if failures. Ask to SHIP when green. Commit, push, handoff. Only on explicit approval. Return to ORIENT after.

Phase transitions: Users can skip phases ("just fix it" = all phases in one pass). After each ACT turn, default to VERIFY. Multiple ACT turns are fine.

Skill Invocation Rule

Before responding in any phase, check if a skill applies. 1% chance it's relevant = invoke it.

Message received → skill applies? → YES: invoke Skill tool first → NO: respond

Priority: Process skills first (brainstorm, systematic-debugging), then implementation skills (task-driven-development, parallel-agents), then quality gates (verification-before-completion, code-review).

Red Flags (you are rationalizing — stop)

• "This is simple" → check anyway
• "I need context first" → skill check comes before exploration
• "The skill is overkill" → if it exists, use it
• "Just this one thing first" → check before acting

Available Skills

Skill	When
godmode:task-driven-development	TDD with serialized task lists and issue chains
godmode:systematic-debugging	Any bug, test failure, unexpected behavior
godmode:brainstorm	Before any creative or design work
godmode:writing-plans	Multi-step task with a spec or requirements
godmode:verification-before-completion	Before claiming work is done
godmode:task-management	Creating, tracking, or executing a task graph
godmode:parallel-agents	2+ independent tasks that can run concurrently
godmode:cap	"cap", "commit and push", "ship it"
godmode:ci-fix	CI failing, "fix CI", broken pipeline
godmode:tackle-issues	Working on GitHub issues in parallel
godmode:code-review	Before merge, after feature complete
godmode:refactoring	Restructuring code without changing behaviour
godmode:receiving-review	Processing incoming review comments
godmode:observability-as-infrastructure	Adding tracing to helpers and subagents
godmode:testing-philosophy	Designing test strategy for new code
godmode:introspection	Auditing skills for consistency after changes
godmode:moa	Multi-model synthesis via Mixture of Agents
godmode:todo-issue-sync	Auditing inline TODOs against tracked issues
godmode:self-reflect	End-of-session retrospective — "reflect"
godmode:wave-integration	Merge parallel agent branches sequentially
godmode:merge	Merge branch, create PR, squash, worktree cleanup
godmode:rust-conventions	Rust coding conventions for writing/reviewing code
godmode:context-map	Map all relevant files before implementation
godmode:decompose	Split large diff/PR into smaller branches
godmode:doublecheck	Three-layer verification of factual claims
godmode:mini-context-graph	Persistent knowledge base with entity graph
godmode:agent-governance	Governance/safety patterns for AI agent systems
godmode:memory-banking	Generate/maintain .ctx/memory-bank/ context
godmode:changelog	Parse git history into structured changelogs
godmode:cross-issue	Cross-repo issue coordination and linking
godmode:dead-code	Find unused public API, orphaned tests, stale refs
godmode:dep-audit	Audit deps via cargo outdated/deny/audit
godmode:dep-bump	Propagate workspace crate version bumps downstream
godmode:doc-maintainer	Audit docs against source code for drift
godmode:health-score	Measure codebase health across seven metrics
godmode:issue-triage	Triage and prioritize GitHub issues
godmode:mistake-tracker	Catalog recurring mistakes and failure modes
godmode:pattern-learner	Cross-session pattern extraction from traces
godmode:pr-author	Compose PR descriptions from branch context
godmode:release-notes	Write user-facing release notes from git history
godmode:workspace-refactor	Catalog breaking changes in shared crate APIs
godmode:agents-skill-save	Create or fix a local skill saved to wrong path
godmode:baml-add-types	Add BAML types/functions to cruxx-agentic
godmode:design	Translate brainstorm into architectural spec
godmode:dual-forge-pr-merge	PR across GitHub and Gitea mirrors
godmode:gh-bulk-issues	Create 3+ GitHub issues with consistent format
godmode:notfiles-release-workflow	Release workflow for the notfiles repo
godmode:planning-with-crux	Design crux DSL pipelines and macro agents
godmode:release-readiness-check	Pre-release verification of tags, crates, gates
godmode:remote-upstream-triage	Fix git push/PR upstream or remote drift
godmode:repo-gap-backlog	Turn local project gaps into GitHub issues
godmode:rust-release-workflow-author	Create GitHub Actions release workflow for Rust
godmode:rustqual	Rust code quality analysis via rustqual CLI
godmode:rustqual-workspace	rustqual guidance for multi-crate Rust workspaces
godmode:session-wrap-commit-push	End-of-session commit and push closeout
godmode:token-cost-optimizer	Analyze or reduce Claude/agent token costs
godmode:using-crux	Navigate, build, or extend the crux codebase
godmode:whatidid	Generate daily activity report from session data
godmode:workspace-bump-commit	Apply version bumps and create release commit
godmode:workspace-release-impact	Decide which crates need version bumps
godmode:writing-solid-rust	SOLID principles and hexagonal arch in Rust
godmode:depgraph	Hexagonal architecture report for Rust workspace
godmode:open-knowledge-discovery	Install and use Open Knowledge on a repository
godmode:open-knowledge-write-skill	Author, draft, and install a new OpenKnowledge skill
godmode:agent-improvement-loop	Collect traces, feedback, evals, HALO diagnosis

Always-Active Rules

• Never --no-verify on commits. Pre-commit hooks always run.
• Commit signing: SSH key via 1Password agent. Unlock 1Password and retry on failure.
• Shell: Nushell (nu) primary. No bash-isms (&&, $(), export VAR=) in .nu files.
• Git in other repos: git -C <path> not cd <path> && git.
• Rust: cargo check + cargo clippy after every change. cargo nextest over cargo test. Fix all clippy warnings before committing.
• Secrets: Never pass raw op:// URIs. Use op read or op run.
• Scope: Touch only files within the explicitly requested scope.
• Destructive ops: Show diff, get explicit confirmation before dropping stashes or deleting files.

Session Ritual

Start:

godmode handon      # triage: running tasks, next runnable, next doob todo

End:

godmode handoff     # warns on leaked running tasks, writes hj handoff

Or with reflection:

godmode handoff && godmode:self-reflect   # handoff + structured retrospective

Additional Resources

• helpers/session-start.sh — run at session start to verify CLI and triage

CLI Quick Reference

godmode handon                          # session start triage
godmode handoff                         # session end closeout
godmode plan ingest <plan.md>           # ingest plan → task graph
godmode task list [--json]              # all tasks
godmode task next [--json]              # next runnable (exit 1 if none)
godmode task add <title> [--id <id>] [opts]  # add task
godmode task start <id>                 # mark running
godmode task done <id> [--commit <sha>] # mark done
godmode task block <id> "<reason>"      # mark blocked
godmode task unblock <id>               # reset to pending
godmode task unblock-all                # reset ALL blocked tasks to pending
godmode task run <id>                   # run task's run: command
godmode task remove <id>               # remove a task from the graph
godmode dispatch [--max 5] [--json]     # parallel chains for orca-strait
godmode agent dispatch <path> [--max N] # ingest + dispatch in one step
godmode status                          # graph counts + next runnable
godmode context [--json]                # session context for hooks/subagents
godmode verify [--crate-name <crate>]   # nextest + clippy + fmt gate