project-understanding

E2E testing for paperlens — scripted gates (bun test), a declarative BDD journey dogfood driven by an agent (browser/device tool resolved per environment), and API drivers. Use when running tests, adding coverage, dogfooding a feature, running a user-story journey, seeding the test account, or executing the manual half of a layer gate.

Autonomous feature implementation in a workspace session. Use when a workspace agent starts building, or when the user says "build", "implement", "execute implementation". Covers the full autonomous flow — initialize harness, implement linearly with one commit per task, review, test, create PR, handle review feedback, and merge. Runs in an isolated git worktree on a feature branch without user interaction.

Pick the next story to work on and bridge it into the feature lifecycle. Use when ready to start building, or when the user says "what's next", "dispatch", "pick a story", "start next feature", "what should I work on". Reads product-context.yaml, syncs workspace signals, computes readiness score for routing, then scores stories by business value + unblock potential + critical path urgency + reuse leverage (penalized by ambiguity + interface risk), assembles context, and hands off to /aep-design or /aep-launch. Supports batch dispatch with WIP limits. For autonomous orchestration, use /aep-autopilot instead.

Generate (or upgrade) a project's e2e-test skill into the BDD layer-gate three-tier shape. Use when setting up E2E testing for a project, scaffolding the e2e-test skill, adding a BDD journey library, migrating a thin/bash e2e-test skill to BDD, or when the user says "scaffold e2e", "set up e2e testing", "add BDD journeys", "upgrade the e2e-test skill". Delegated to by /aep-scaffold. Emits a canonical, cross-tool skill (Claude Code + Codex + Pi) with a separate test-tool-selection guide.

Product-level opportunity and product framing. Use when starting a new product from scratch, revisiting product direction, or when the user says "new product idea", "validate this idea", "product framing", "what should we build", "revisit our assumptions". Walks through opportunity validation and produces a Context Document that feeds into /aep-map and /aep-design.

Host-agnostic executor abstraction for spawning and steering implementation agents. This is a utility skill — it defines how /aep-launch, /aep-build, and /aep-autopilot start a worktree-bound agent, send it mid-flight instructions, check liveness, surface human gates, present it for review, and tear it down, across Claude Code (native background subagents / background sessions) or Codex (native subagents / exec workers), with tmux+cmux as the legacy pinned fallback. Reference its files from any skill that needs to run work in an isolated workspace. Triggers on "executor", "which backend", "launch mode", "spawn workspace", "run under codex", "native-bg-subagent", "agent teams", "claude-team", "no tmux", "with tmux", "run as a workflow", "host detection".

Reusable generator/evaluator pattern for honest artifact validation. This is a utility skill — it provides the scoring framework, agent contracts, evaluation protocol, and findings format used by /aep-build, /aep-validate, and any skill that needs to evaluate agent-produced work. Use directly when you need to run a gen/eval loop on any artifact, or reference its files from other skills. Triggers on "gen/eval", "generator evaluator", "evaluate honestly", "separate evaluator", "scoring framework".

Spawn an autonomous workspace agent for feature implementation. Use after /aep-design is complete, or when the user says "launch workspace", "start building", "spawn agent", "send it to build". Creates a git worktree on a feature branch, starts the selected executor mode (Claude Code native background subagents or background sessions; Codex native subagents or exec workers; tmux only when pinned), and optionally sets up a separate evaluator agent. Followed by /aep-build (which runs autonomously in the workspace).

Decompose a product into system map, layered story graph, and agent topology. Use after /aep-envision, or when the user says "decompose", "story map", "system architecture", "break this down", "plan the stories", "agent topology". Produces the complete execution plan that the feature workflow operates on.

Object-first UI structure modeling (OOUX/ORCA) for UI-facing products. Use after /aep-map and before dispatching UI-facing stories, or when the user says "object model", "object map", "objectify", "OOUX", "ORCA", "noun-first IA", "what objects/screens". Mines a draft Object Map (objects → relationships → CTAs → attributes → screens) from the story map, presents it for a short human review gate, and on approval writes the noun-first blueprint that build agents follow. Bridges the verb-first story map to the UI so agents stop inventing task-wizard screens per story.

Classify feedback and update product context after shipping features. Use after /aep-wrap, after user testing, or when the user says "reflect", "what did we learn", "update the product context", "classify feedback", "replan". Closes the feedback loop by routing observations back to the right phase.

Scaffold a new project or onboard/converge an existing one with agentic development infrastructure. Use when creating a new project ("new project", "scaffold", "create app") OR setting up / repairing an existing project ("onboard project", "set up existing project", "initialize infrastructure", "add workflow to project", "fix the skills layout", "upgrade to canonical"). For new projects, creates a full-stack monorepo via Better-T-Stack. For existing projects, runs an idempotent audit → confirm → converge that repairs drift (canonical cross-tool skills layout, BDD e2e skill, pin). Both modes set up OpenSpec, workspace hook, and the BDD e2e-test skill.

Generator/evaluator validation for any AEP artifact — product context, architecture, stories, code, or documents. Use after any generation phase (/aep-envision, /aep-map, /aep-design, /aep-build) or when the user says "validate", "verify", "check the design", "dry-run", "evaluate", "gen/eval", "generator evaluator". Spawns parallel agents: a Generator (dry-run the artifact), an Evaluator (check against reality), and optionally a Protocol Checker (verify downstream compatibility). Modifies only the artifact being validated — never implements code.

Continuously ingest bug trackers, error streams, and telemetry; classify each finding with the /aep-reflect classifier; dedupe against existing stories; and auto-create bug/refinement stories into product-context.yaml so reflect→dispatch becomes a self-feeding loop. Use when the user says "watch", "monitor for new work", "ingest errors", "auto-create stories from telemetry", "keep an eye on the bug tracker", "/aep-watch", or wants new work to enter the backlog without manually running /aep-envision or /aep-reflect. Runs from the MAIN workspace only.

Capture and route workflow learnings between downstream projects and AEP. Use after /aep-wrap or /aep-reflect when process observations need to be standardized, when reviewing what a build run taught about AEP workflows, or when pulling learnings from downstream projects back upstream. Triggers on "workflow feedback", "capture learnings", "what did we learn about process", "pull learnings from downstreams", "upstream lessons".

Reusable pattern for authoring a dynamic workflow — a custom multi-agent harness Claude writes on the fly for one task, run as a Claude Code Workflow (a deterministic JS script that spawns and coordinates context-isolated subagents). This is a utility skill — it provides the sub-pattern catalog (classify-and-route, fan-out-and-synthesize, adversarial verification, generate-and-filter, tournament, loop-until-done) and the "when to reach for a workflow" judgment. Cross-linked from /aep-gen-eval and /aep-executor; relevant when /aep-dispatch or /aep-validate work would benefit from a per-task harness. Use directly when a task is large, uncertain, needs adversarial verification, or runs at scale and the default single-context harness would be lazy, biased, or drift off-goal. Triggers on "dynamic workflow", "ultracode", "write a harness", "harness for this task", "…with workflow", "orchestrate subagents". NOT for capturing process feedback — that is /aep-workflow-feedback.

Post-merge archive and cleanup. Use after a PR has been merged, or when the user says "wrap up", "archive", "cleanup after merge", "post-merge". Runs OpenSpec archive, commits the archive, and removes the workspace. The final step in the feature lifecycle.

Orchestrate the full dispatch-launch-monitor-review-wrap-dispatch cycle autonomously. One command to go hands-free. Use when the user says "autopilot", "run autonomously", "auto dispatch loop", "hands-free mode", "start building everything", "go auto", "run the pipeline", "let it run", "manage workspaces", or wants to dispatch and monitor multiple stories without manual intervention. Always trigger this over /aep-dispatch when the user wants continuous autonomous operation rather than a single story dispatch. Runs from the main workspace only.

Full environment onboarding and first-hour orientation for Agentic Engineering Patterns (AEP). Use when setting up a new machine, joining the project, installing the plugin, or when the user says "get started", "onboard", "setup environment", "install prerequisites", "what is this plugin", "help me understand AEP", "I'm new to agentic engineering", "introduce me to this workflow", or wants to prepare their dev environment AND learn the mental model. Always use this first when a user mentions AEP, agentic-engineering-patterns, or this plugin for the first time.

Human alignment checkpoint for any quality dimension during rapid agent development. Use when /aep-reflect identifies a gap between "works" and "right", when the user says "calibrate", "align", "design brief", "capture", or before dispatching .5 alignment layers. Supports calibration types — visual-design, ux-flow, api-surface, data-model, scope-direction, copy-tone, performance-quality. Phase 1 generates a dimension-specific brief; Phase 2 captures decisions into the appropriate artifact.

Interactive feature design on the integration branch (main, or develop in two-branch mode). Use when starting a new feature, or when the user says "design a feature", "let's design", "explore and propose". Runs OpenSpec explore, propose, and design review phases interactively with the user, then commits artifacts to the integration branch. The first step in the feature lifecycle — followed by /aep-launch.

AEP-specific reference for git + git worktree workflows. Use when the user asks "how do I create a worktree?", "what's the AEP branch convention?", "how do I clean up a worktree?", "how does AEP use git?", "remind me of git commands for parallel agents", or needs to recover from a worktree mishap. Documents worktree lifecycle, branch naming, the one-commit-per-task pattern, recovery procedures, and PR conventions used by `/aep-launch`, `/aep-build`, and `/aep-wrap`.

Forge reusable skills from accumulated project-memory lessons. Use when the user wants to consolidate, distill, or extract durable skills out of project-memory/ — turning lessons-learned into agent-loadable skills that prevent the project from repeating its own mistakes. Trigger phrases: 'forge a skill', 'forge memory', 'distill lessons', 'extract skills from memory', 'consolidate project-memory', 'turn lessons into skills', 'evolve our skills', 'memory curator', 'what skill should this become'. Trigger proactively when project-memory wrap-up has produced 3+ new lessons since the last forge, when the user is about to open a PR after several captured lessons, or when a recurring theme emerges across 2+ session files. Reads project-memory/lesson-learned/ (and other subsystems), clusters lessons ≥7 days old, then merges them into umbrella skills under the host project's detected skill-loading path.

Author or update a project's AGENTS.md to install behavioral guidelines that govern any agent acting in that project. Use when the user wants to configure project-level agent behavior, set up or extend AGENTS.md, add a behavioral floor, prepend a coding-discipline preamble, or apply a named behavior pack to a project. Bundles the Karpathy behavioral preamble as one reference pack and supports adding more. Trigger phrases include "set up AGENTS.md", "configure agent behavior", "add behavioral guidelines", "add karpathy preamble", "install behavior pack", "create AGENTS.md", "control how agents behave in this repo".

Use when the user wants to capture, organize, or retrieve project knowledge as a git-committable memory system at project-memory/ at the project root. Trigger proactively on phrases: 'lesson learned', 'save lesson', 'wrap up session', 'postmortem', 'retro', 'capture this', 'what did we learn', 'why did this fail', 'have we seen this before', 'add to lesson', 'end of session', 'project memory', 'team memory'. Also trigger when a mission is ending and no lesson has been captured yet, or when the user asks a question whose answer likely lives in prior work. Backed by qmd for hybrid BM25 + vector search over every subsystem under project-memory/.