AgenticPW
UserA practical collection of agentic practices and workflows, including reusable skills, commands, reference notes, and base knowledge documents.
Categories
Indexed Skills (17)
apw-brainstorm
Structured divergent-then-convergent ideation workflow. Use when asked to brainstorm, ideate, generate options, or explore approaches before narrowing to a decision.
apw-build-loop
Build a new orchestrator command that loops a user-chosen set of phases (e.g. research + implement + review). Use when asked to create a new apw-* workflow command from a set of phases — the skill resolves the phases to subagents, asks about the loop shape, confirms, then writes the command (and any missing subagents) following the house conventions.
apw-commit
Analyze pending git changes, draft a commit message using a project schema, and commit/push after user review.
apw-decompose
Structured task-decomposition workflow. Use when a task needs to be mapped to an ordered sequence of skills before any work begins — to decide the right workflow rather than jumping straight to implementation or reflexively running the full research→design→plan→implement→verify→review stack.
apw-find-root-cause
Structured root-cause diagnosis workflow. Use when a known defect or failure must be diagnosed and its cause proven before anyone attempts a fix.
apw-gather-requirements
Interactive requirements-gathering workflow. Use when a task's requirements must be drawn out through iterative questioning before a requirements.md doc is written.
apw-handoff
Checkpoint in-flight task state to the work item, and resume from that checkpoint in a fresh session. Use when a session is ending mid-task, when context is about to be compacted or reset, or when picking up a task a previous session left unfinished.
apw-implement
Structured implementation workflow. Use when asked to implement a feature, fix, or any change.
apw-learn
Distill durable lessons from a conversation, user input, or your own work into the project's AGENTS.md — but only after the user confirms. Use after finishing a task, when the user wants to record a rule, or when you've figured out something worth persisting.
apw-onboard
Build and maintain a durable codebase-orientation record in AGENTS.md so sessions stop re-deriving the codebase from scratch, and orient from it by verifying against reality first. Use when explicitly asked to onboard, to build/refresh the orientation, or to get oriented in a codebase.
apw-plan
Structured planning workflow. Use when a task must be turned into a clear, checkable plan — optionally with requirements and design docs — before implementation begins.
apw-research
Structured research workflow. Use when a question must be investigated before it can be implemented or decided.
apw-review
Structured adversarial self-review workflow. Use when about to ship your own diff, plan, or conclusion and you want to find what's wrong with it before someone else does.
apw-tech-design
Structured technical design workflow. Use when a task needs a concrete technical solution design — options, tradeoffs, and a recommendation — before implementation begins.
apw-verify
Structured verification workflow. Use when asked to verify that an implementation meets its requirements, spec, or plan, and to record the result.
apw-write-test
Structured test-writing workflow. Use when asked to write new tests for a unit of behavior — in TDD (tests first) or post-hoc (tests after) mode.
apw-work-with-work-item-artifacts
Defines where a work item's artifacts live. Use as the single source of truth for the location of workflow, research, brainstorm, requirements, design, plan, root-cause, verification, review, handoff and other "apw-skills" documents.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.