AI Agent Workflow Collection

Use when the user asks to "scaffold a new ycc skill", "add a skill to the bundle", "extend ycc", "create a new command under ycc", "add an agent to ycc", or when the user wants to author new source-of-truth content in the ycc/ bundle (skills, commands, agents). Contributor workflow — not a generic plugin scaffolder. Respects the source-of-truth / regenerated-bundle split.

Use when the user asks to "release ycc", "bump ycc version", "prepare a ycc release", "cut a new ycc release", "tag a ycc release", or when the user wants to version, regenerate, validate, and produce release notes for the ycc bundle. Orchestrates existing generators/validators; never auto-commits.

Generate structured implementation reports documenting changes made during plan execution. Use as optional Step 4 after implement-plan to create reports in docs/plans/[feature-name]/report.md with overview, files changed, features, and test guidance.

Use when the user asks to "audit ycc compatibility", "check Cursor/Codex bundle health", "verify cross-target parity", "run a compatibility report for ycc", "is ycc ready to release", "are the generated bundles up to date", "check if the bundles are in sync", "validate generated targets", "are Cursor and Codex bundles current", "check bundle drift", "report on target feature gaps", or when the user wants a structured per-target health report covering drift detection, install-assumption validation, and feature-capability gaps across the Claude, Cursor, Codex, and opencode targets without bumping versions or committing anything.

Use when the user asks to "initialize formatters", "bootstrap linting", "add lint and format scripts", "set up prettier / biome / ruff / clippy / golangci-lint / shellcheck", "add a Makefile lint target", "install a style bundle", "wire a pre-commit hook for linting", or any request to stand up a best-practices lint/format environment in a project. Profiles the target, installs a self-contained scripts/style.sh bundle, configures per-language tool files, injects runnable aliases (package.json / Makefile / justfile), appends a "Linting & Formatting" section to the project docs, and optionally wires CI and pre-commit hooks. Idempotent via a managed manifest — safe to re-run with --sync.

Convert rule-file hook guidance into target-appropriate config. Reads hook recommendations from the resolved language rules, consults the target-capability matrix, and emits only config the target can execute (Claude → JSON; Cursor → advisory rule embedding where supported; Codex → advisory-only). Use when the user asks to "generate hook config", "configure Claude hooks from ycc rules", "apply ycc hooks", "cross-target hook setup", "hook matrix for this repo", or "verify my hook config".

Create detailed parallel implementation plans by running analysis and validation stages, then synthesizing dependency-aware tasks into parallel-plan.md. Use after shared-context to prepare implementation-ready planning artifacts. Defaults to standalone parallel sub-agents via Cursor; pass --team (Claude Code only) to orchestrate the analysis and validation stages as teammates under a shared TeamCreate/TaskList with coordinated shutdown.

Lightweight natural-language git commit helper. Interprets the argument as either a glob, a filter phrase ("except tests", "only new files"), a topic phrase ("the auth changes"), or leaves staging alone if the user says "staged". Stages the matching files and generates a conventional commit message. Use for quick single-purpose commits; use /git-workflow instead when you need documentation updates, parallel agents, or full PR orchestration. Adapted from PRPs-agentic-eng by Wirasm.

Execute a PRP plan file with per-task validation loops. Detects package manager, prepares git branch, runs 5 validation levels (static/unit/build/integration/edge), writes a report to docs/prps/reports/, and archives the plan. Auto-detects parallel-capable plans and prompts for sequential or parallel execution. Use when the user asks to "execute a PRP plan", "implement from a plan file", "run prp-implement", "parallel PRP implement", "team PRP implement", or provides a path to a .plan.md file.

Create a self-contained feature implementation plan with codebase pattern extraction and optional external research. Detects PRD vs free-form input, runs codebase discovery via prp-researcher, and writes docs/prps/plans/{name}.plan.md ready for /prp-implement. Use when the user asks for a "PRP plan", "implementation plan from PRD", "feature plan with patterns to mirror", "parallel PRP plan", "team PRP plan", or says "/prp-plan".

Create a GitHub pull request from the current branch — validates preconditions, discovers PR templates, analyzes commits and file diffs, references PRP artifacts (prds/plans/reports), pushes, and creates the PR via gh. Lightweight sibling of /git-workflow --pr; use this when you only need the PR, not the commit+docs orchestration. Adapted from PRPs-agentic-eng by Wirasm.

Generate a lightweight feature spec for the PRP workflow — single-pass with optional codebase/market grounding via prp-researcher. Produces a concise spec at docs/prps/specs/{name}.spec.md covering problem statement, requirements, technical approach, integration points, and risks. Accepts a free-form description or starts with clarifying questions. Parallel entry point to prp-prd — use prp-spec when the problem is clear, prp-prd when it needs discovery. Both feed directly into prp-plan. Use when the user asks to "write a spec", "spec out a feature", "quick spec", "create a PRP spec", or says "/prp-spec". For heavyweight multi-agent research, use feature-research instead.

Use when the user asks to "prepare a release", "cut a release", "tag a release", "create a GitHub release", "draft release notes", "set up release CI", "add a release workflow", "audit release pipeline", "publish a GitHub release", or "publish release notes" for any project. Detects language/toolchain, drafts a changelog, plans platform/arch artifacts, optionally generates or audits GitHub Actions release CI, and emits the exact commands the user runs to tag and publish. Never auto-commits, pushes, or publishes without an explicit --publish and --confirm from the user. Generic external-project counterpart to bundle-release (which is scoped to this repo's ycc bundle).

Use when the user asks to "create issues from research", "turn research into GitHub issues", "create issues from feature spec", "turn plan into issues", "issues from parallel plan", "create issues from PRP plan", "generate issues from deep-research output", "create tracking issues from plan", "parse plan into actionable issues", or mentions converting documentation produced by deep-research, feature-research, parallel-plan, plan-workflow, or prp-plan into structured GitHub issues with labels and tracking hierarchy.

Capture the current session state — what was built, what worked, what failed, open questions, and exact next step — to a dated file at ~/.cursor/session-data/YYYY-MM-DD-{shortid}-session.tmp so a future session can resume with full context. Use when the user says "save session", "end of session", "preserve context", "before context limit", or says "/save-session". The file is meant to be read by /resume-session at the start of the next session.

Create shared context documentation for a feature by deploying 4 researchers (architecture, patterns, integration, docs) that write research artifacts, then synthesizing verified findings into shared.md. Use as Step 1 before parallel-plan when preparing implementation context. Default is standalone parallel sub-agents via Cursor. Pass --team (Claude Code only) to deploy the 4 researchers as teammates under a shared TeamCreate/TaskList with coordinated shutdown.

Idiomatic TypeScript patterns — strict type system, discriminated unions, generic inference, satisfies, branded types, errors as values, ESM/CJS modules with exports maps, Promise combinators and AbortController, iterators and higher-order functions, runtime selection across Node/Deno/Bun/browser/edge, and the Vite + Vitest toolchain stack. Use when the user is writing new TypeScript, reviewing TS code, refactoring, designing libraries for npm, tuning tsconfig strict mode, asks about generics/conditional types/discriminated unions/branded types/satisfies, asks about ESM/CJS dual publishing and exports maps, asks about Promise combinators/AbortController/async iterators, asks about Vite config or vite build --lib, chooses between Node/Deno/Bun/browser/edge runtimes, or wants TypeScript idioms and anti-patterns.