Skip to content
Development
Skill

/no-diary

Cut slop, jargon, process narration, past states, session leakage, and provenance from a deliverable (PR or MR body, code comment, doc, skill, review comment, issue, plan) so it states the result rather than how the work happened. Use when asked to de-slop, de-jargon, trim,

From plugin
claude
1882 skills12 agents1 MCP

Claude Code

$ npx -y skills add bendrucker/claude --skill no-diary --agent claude-code

SKILL.md

no-diary.SKILL.md
name: writing:no-diary
description: >-
  Cut slop, jargon, process narration, past states, session leakage, and
  provenance from a deliverable (PR or MR body, code comment, doc, skill, review
  comment, issue, plan) so it states the result rather than how the work
  happened. Use when asked to de-slop, de-jargon, trim, shorten, or cut down one
  of those, or when told that one reads as slop or runs too long.
argument-hint: "[<file> | <pr> | <section> | <text>]"
user-invocable: true
allowed-tools:
  - Read
  - Edit
  - Write
  - Grep
  - Bash
  - Agent
  - Skill

No Diary

Arguments

`$ARGUMENTS` is freeform. Resolve it to concrete text:

  • A path, optionally with a line range, reads that file.
  • A PR or MR number or URL reads the body with the commands in `references/surfaces.md`.
  • A section name or heading locates it in the artifact under discussion.
  • A description with no target names the problem in the deliverable in play.
  • A fraction or a keep-list (`50%`, `keep: the warning and the upstream link`) sets the target length. Cut to it, taking the `Removals` in order, and say what went beyond the diary.

With no arguments, resolve the target in this order: text the user pasted this turn, the file this session last wrote or edited, then the PR for the current branch. Ask only when two candidates are equally live.

Output

  • **A file.** Edit it in place.
  • **A PR or MR body.** Write the new body under `tmp/`, then apply it with the command in `references/surfaces.md`. Confirm first when the user did not name the PR.
  • **Pasted text.** Return the rewrite in your reply and leave the disk untouched.

Then report what you cut, grouped by the `Removals` headings, in a few lines.

The Test

Every sentence changes what the reader does. A sentence explaining why a decision was right, what it replaced, or where it came from fails.

Delete whole sentences. Cut within a sentence only for the cases in `Gotchas`.

Judge the text as a reader who was not in the session. For a whole artifact rather than a passage, dispatch a `general-purpose` agent that loads `writing:no-diary` and `writing:writing` and works from the artifact alone, plus its diff where one exists, with no session transcript.

The rewrite is done when every sentence passes the test and every fact a reader needs from the original survives.

Removals

  • **Change evolution.** The shapes the work passed through. Any sentence needing an earlier draft to parse, and sequencing words such as "originally", "initially", "then", "turned out", "along the way", "ended up".
  • **Session leakage.** The conversation inside the artifact: feedback echoed as commentary, a reviewer's question answered in a code comment, the user's own phrasing quoted back.
  • **Past states.** What the code or doc used to be, and how bad it was. Remove the thing rather than commenting on its removal.
  • **Provenance and deliberation.** Where an idea came from, what it beat, why a rule is correct. Alternatives belong in the commit or the PR body, so when the artifact is a PR body, keep a decision's reason in one clause.
  • **Restatement.** Prose that re-describes the diff: file inventories, structural tours, a paragraph per function. Name the function instead.
  • **Self-reference.** The artifact describing itself, its own sections, or harness behavior the harness already provides.

Replacements

Most diary sentences sit where a useful sentence belongs. Put one of these there:

  • The end state, stated plainly.
  • What a reader cannot reconstruct from the diff: intent, the decisions behind it, what a change prevents.
  • How, only when it is novel.
  • Insight earned along the way, stated briefly as its own point.
  • Evidence, such as a number proving an assumption held.

Gotchas

  • A prompt asking for what is non-obvious to a reviewer produces the literal heading "Things that wouldn't be obvious to a reviewer". Write what the instruction asked for, never the instruction.
  • Verification results resemble narration. Keep the result and cut the framing: a row count proving a join assumption is a finding, and "I ran it and found" is not. A passing-test tally is not a result: cut "all tests pass" and test counts.
  • Cutting a rationale can strand the rule it supported. Keep one clause when the reason is what lets a reader handle an unlisted case.
  • Rewriting a user's own prose is out of scope. When the invocation frames the text as theirs ("my review comment", "my draft") or the artifact attributes it to them, cut the diary and leave the voice.
  • A review comment names the code by function or behavior. Drop `file:line` paths and finding counts ("three nits") from it, the user's own comment included.

Surfaces

[`references/surfaces.md`](references/surfaces.md) has rules per surface. Read the section for the surface you are editing.

Voice

Load `writing:writing` before rewriting. It governs word choice and cadence.

Read more
@bendrucker
Ships withclaude

My personal plugin marketplace for Claude Code, Anthropic's AI coding assistant.

1882 skills12 agents1 MCP
Get the whole plugin

Stars on GitHub

Momentum

stars in total
18
Star history of bendrucker/claude on GitHub, a chart by star-history.com

Other skills on claude.

cli

cli

Type-safe CLI argument parsing with commander and @commander-js/extra-typings, the standard parser for this repo's Bun scripts.…

@bendrucker@bendruckerInstall Skill
prose-lab

prose-lab

Turn the user's labels on repo prose, or their notes from reviewing a document in Plannotator, into calibrated Vale rules.

@bendrucker@bendruckerInstall Skill
activity

activity

Report real device usage from ActivityWatch. Covers per-app time, window titles, and active vs idle spans. Use when asked "what…

@bendrucker@bendruckerInstall Skill
agent-team

agent-team

Orchestrating Claude Code agent teams. Use when creating teams, spawning teammates, assigning tasks, configuring teammate modes,…

@bendrucker@bendruckerInstall Skill
handoff

handoff

Hand the current conversation off to a fresh background agent that picks up the work immediately, so you can step away and track…

@bendrucker@bendruckerInstall Skill

More skills from other plugins.

brainstorming
Auto-invoked

brainstorming

You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or…

@obra@obraInstall Skill
ask-matt

ask-matt

Ask which skill or flow fits your situation. A router over the skills in this repo.

skills11d309k
@mattpocock@mattpocockInstall Skill
accessibility
Auto-invoked

accessibility

Design, implement, and audit accessible UI to WCAG 2.2 Level AA across Web, iOS, and Android — semantic ARIA roles and labels,…

ecc13d293k
@affaan-m@affaan-mInstall Skill
compiler-commit

compiler-commit

Use when you want to verify compiler changes and commit with the correct convention. Runs tests, lint, and format, then commits…

react2d252k
@facebook@facebookInstall Skill

Questions about no-diary.

Which coding agents can use no-diary?
no-diary installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.