prompt-evaluation-runn…
Use when evaluating prompts, LLM outputs, red-team suites, or model behavior with local eval…
Create a concise Product Development Requirement and one or more implementation-sized YYLO Ledger tasks when the user explicitly asks to plan or register work. Planning-only requests stay as external drafts; Ledger records and tasks are created only when the user explicitly asks
$ npx -y skills add yeaight7/agent-powerups --skill plan-ledger-tasks-yylo --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/plan-ledger-tasks-yyloContext preview
The summary Claude sees to decide when to auto-load this skill.
Create a concise Product Development Requirement and one or more implementation-sized YYLO Ledger tasks when the user explicitly asks to plan or register work. Planning-only requests stay as external drafts; Ledger records and tasks are created only when the user explicitly asks
name: plan-ledger-tasks-yylo description: Create a concise Product Development Requirement and one or more implementation-sized YYLO Ledger tasks when the user explicitly asks to plan or register work. Planning-only requests stay as external drafts; Ledger records and tasks are created only when the user explicitly asks to register the work or confirms when asked.
Turn a requested change into one concise Product Development Requirement (PDR) and one or more implementation-sized YYLO Ledger tasks, with the PDR captured as an immutable artifact receipt. Planning and registration are separate steps: a plan is drafted first, and persistent Ledger records are written only on an explicit request to register.
A planning-only request ("plan this", "break this down") produces the draft package only — the external PDR draft plus a proposed task split. No Ledger record, task, or receipt is created until the user explicitly asks to register the work in the Ledger (or confirms when asked).
Do not use when:
Required tools (both):
`yy ledger` delegates every Ledger command to the separately installed `yylo-ledger` executable; it is not bundled with `@yylo/cli`. `yy --version` passing therefore does not prove the Ledger workflow can run — check both commands below before promising any Ledger work.
Check:
yy --version yylo-ledger --version
Install (requires user approval — do not auto-install):
npm install -g @yylo/cli python -m pip install yylo-ledger
For exact-version installs, the YYLO CLI documentation pins the Ledger package (for example `python -m pip install 'yylo-ledger==0.4.0'`, matching the Ledger compatibility policy stated by that `yy` release).
Fallback:
1. Read the project instructions and the relevant product code. Read existing task and spec metadata through the installed `yy ledger` commands; do not assume a specific plan file exists. 2. Draft one concise PDR covering goal, current behavior, scope, exclusions, risks, dependencies, acceptance criteria, and focused tests. Keep the draft in a fresh external file — not inside the product tree and not inside a task body. 3. Draft the task split next, still without writing anything: split work into tasks only where pieces can be implemented and validated independently, and note explicit path ownership and dependencies for concurrent tasks. A planning-only request ends here — present the external PDR draft plus the proposed task split, and offer registration as an explicit next step. 4. Register only on an explicit request: create any Ledger record, task, or receipt only when the user explicitly asked to register the work in the Ledger, or explicitly confirms when asked. If the request was planning-only and the user declines or does not answer, deliver the external draft and stop — never create persistent records, tasks, or receipts unprompted. 5. Preflight `yy ledger --help` and `yy ledger artifact --help`. Capture the PDR as a local immutable `report` Artifact Record with task/request provenance, then verify its ID, digest, size, retention, retrieval, and history. 6. Create tasks through routed `yy ledger` commands. Put concise durable requirements and acceptance criteria in each task body, record the PDR artifact ID in supported task fields or provenance, and relate follow-ups instead of reopening archived task IDs. 7. Keep planning metadata out of product worktrees: a product repository ships product documentation only. 8. Do not start implementation, create worktrees, push, deploy, or mutate production unless the user separately asks.
Use `--id`, not legacy `--ID`, for task mutations. Return the task IDs and a short dependency/order summary.
Curated power-ups for coding agents: skills, slash commands, MCP configs, hooks, AGENTS.md templates, and workflows for serious software engineering. Claude Code, Codex, Antigravity CLI, Cursor and more
Repo: yeaight7/agent-powerups
Use when evaluating prompts, LLM outputs, red-team suites, or model behavior with local eval…
Use when creating or reviewing red-team eval plugins, attack templates, grader rubrics,…
Use when designing, running, debugging, or hardening deterministic eval suites for agent…
Use when designing tool definitions for a new agent or subagent, an agent shows high retry…
Use when routing a prompt to a local provider CLI for a second opinion, review, or plan --…
Use when starting work in an unfamiliar area of a codebase, spawning a subagent that needs…