agent-doc-discipline
Writing-time discipline for documents agents consume (the five surfaces, specs, tickets, .omc/skills/) — every rule checkable and…
Aim a vague task at its real target without overspecifying: name the consumer, the observable change, the done-check, and the non-goals; at most one question when a real fork remains.
Claude Code
$ npx -y skills add Yeachan-Heo/oh-my-claudecode --skill sharpen --agent claude-code
Codex
$ npx skills add Yeachan-Heo/oh-my-claudecode --skill sharpen -a codex -y
Cursor
$ npx skills add Yeachan-Heo/oh-my-claudecode --skill sharpen -a cursor -y
Gemini CLI
$ npx skills add Yeachan-Heo/oh-my-claudecode --skill sharpen -a gemini-cli -y
GitHub Copilot
$ npx skills add Yeachan-Heo/oh-my-claudecode --skill sharpen -a github-copilot -y
OpenCode
$ npx skills add Yeachan-Heo/oh-my-claudecode --skill sharpen -a opencode -y
Windsurf
$ npx skills add Yeachan-Heo/oh-my-claudecode --skill sharpen -a windsurf -y
name: sharpen
description: "Aim a vague task at its real target without overspecifying: name the consumer, the observable change, the done-check, and the non-goals; at most one question when a real fork remains."
aliases: []
metadata:
credits:
task-aiming: pstack
author: Lauren Tan
url: "https://github.com/cursor/plugins/tree/main/pstack"
port: pstack-claude
port-author: Michael Denyer
port-url: "https://github.com/michael-denyer/pstack-claude"A vague brief fails quietly. Sharpening names what the work is for so the first attempt lands near the target — without pinning details the work itself should decide.
1. **Name the consumer**: who feels the result (an end user, a teammate, a future reader), and what changes for them, observably. 2. **Run the expert lens** — the goal is a deeper task, not a bigger one. Before rewriting the brief, answer four questions and fold the answers in:
does not ask for
already been digested for the session
right — always keep one such outside check 3. **Name the done-check**: the cheapest observable that says the task is done — a command that passes, a page that renders, a question that answers. 4. **Name the non-goals**: what is explicitly out of scope this round. Non-goals are the cheapest scope protection there is. 5. **One question, if a real fork remains**: if two different tasks both satisfy the brief and the choice changes the work materially, ask — one question, with the options and what each would cost. Otherwise proceed and report the interpretation you took (`principles` skill, Never block on the human).
The deliverable is the rewritten brief, deeper not bigger, with choices that are the user's left open instead of decided.
user pinned it.
execute without chat context, in three sentences or fewer.
findable), say so and route to `deep-interview`.
For Codex users: Check out oh-my-codex — the same orchestration experience for OpenAI Codex CLI. Liked OmC but found it a bit overkill? Try gajae-code.
Writing-time discipline for documents agents consume (the five surfaces, specs, tickets, .omc/skills/) — every rule checkable and…
Clean AI-generated code slop with a regression-safe, deletion-first workflow and optional reviewer-only mode
Settle types and module shape before cross-boundary code — competing designs explored in parallel, compared on API surface and…
Periodic architecture survey — walks the module graph and reports ranked deepening candidates (shallow modules, hypothetical…
Run N parallel candidates at one brief and synthesize: rubric-first framing, isolated fan-out, a cross-judge on a different…
Shipyard's navigator — chart a foggy effort (destination unclear, questions not yet stateable) into a map of decision tickets on…
Use your assigned AgentMail inbox to read email tasks, explicitly send or reply, and check delivery. Provided automatically by…
Create a new Architecture Decision Record with sequential numbering and AgentDB registration