agentleFS
Agents4Academia-AIAgents4Academia-AI/UReKA/.claude/skills/edit/SKILL.md

edit

Skill9 starsChanged 5 months agoMITView on GitHub

Co-edit a paper or concept page in place — action inline review comments, follow chat instructions, or build on the user's own edits — grounded in the page's cited sources, preserving frontmatter and wikilinks

18 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use daily-arxiv, exp-eval and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • LambenthanSkill

    daily-arxiv

    每日从 arXiv 拉取新论文,过滤相关性,auto-ingest 高优先级论文,检测 SOTA 更新

    Also in 5 projects that use this

  • LambenthanSkill

    exp-eval

    实验判决门:Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges

    Also in 5 projects that use this

  • skyllwtSkill

    exp-pilot-eval

    Pilot result evaluation — read pilot results, apply success criteria, update idea page (pilot_result, failure_reason if failed), generate PILOT_VERDICT_REPORT. Called by /ideate Phase 5 after /exp-pilot-run.

    Also in 4 projects that use this

  • skyllwtSkill

    exp-pilot-run

    Pilot experiment execution — read Pilot Spec YAML, write pilot code, run experiment(Confirm with the user before operation and require the applicant to conduct manual inspection), return results. Called by /ideate Phase 5. Does NOT modify wiki pages or judge pass/fail.

    Also in 4 projects that use this

  • phanghonghaoSkill

    poster

    HTML/CSS A4 海报生成。从沙龙/活动 Markdown 文件生成科技风 A4 海报 PDF/PNG。Use when user mentions: poster, 海报, 海报生成, make poster, generate poster, 活动海报, 沙龙海报.

    Also in 4 projects that use this

  • +6 more from Lambenthan
  • +1 more from skyllwt

The file

We split this skill into the rules it gives an agent

8 rules. 0 are common, 6 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 16 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If $ARGUMENTS is empty, ask the user which page to work on, then continue.

    line 21
  2. Very specific:

    Load its sources: Read the page's sources: frontmatter list — these are the files it was synthesised from.

    line 28
  3. Distinctive:

    If a request needs information not present in the sources, do not invent it.

    line 44
  4. Distinctive:

    If a request is ambiguous, ask the user instead of guessing.

    line 48
  5. Distinctive:

    Tutor re-test reminder: After confirming, check whether the edited page belongs to a learning course and, if so, respect the user's chosen …

    line 54
  6. Distinctive:

    Preserve Obsidian/Markdown syntax exactly: Treat these as literal content — never reformat, normalise, or rewrite them unless a request …

    line 64
  7. Very specific:

    Write math in LaTeX: When a request adds or revises math, use LaTeX (inline $ ...

    line 72
  8. Distinctive:

    After a successful edit, check whether the page is covered by a learning course and remind the user to re-test at their chosen cadence:

    line 85

“Common” is measured: kept by at least 6 in 10 of the 16 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Co-edit a knowledge-base page with the user, editing it in place. This is the collaborative refinement step for a papers/ or concepts/ page: the user drives and you make the changes. Any factual content must come only from the sources the page was synthesised from — if the sources don't cover what's asked, point out the gap instead of inventing it.

The user works in three (combinable) modes:

  • Inline comments — they mark requests addressed to you inline in the file, e.g. @claude: phrase this more precisely and less bombastically.
  • Chat — they tell you in conversation what to change ("tighten the overview", "add a limitations paragraph from the InstructGPT source").
  • Direct edits — they've already edited the page themselves; re-read it so your changes build on their current text, and never revert their wording.

If $ARGUMENTS is empty, ask the user which page to work on, then continue.

Steps

  1. Read the page. Read the file at $ARGUMENTS (always re-read so you see the user's latest direct edits).
  2. Load its sources. Read the page's sources: frontmatter list — these are the files it was synthesised from. Resolve each entry relative to the page; if that path doesn't exist, locate the file by its basename with Glob (e.g. sources/**/<name>.md, notes/**/<name>.md). Read every resolved source. These are the only admissible basis for new or changed factual content.
  3. Gather the requests. Collect both the @claude: ... comments in the page (inline or on their own line; each refers to the text around it unless it says otherwise) and any instructions the user gave in chat. Comments without the @claude: marker are the user's private notes — leave them untouched. If there are no comments and no chat instructions, ask the user what they'd like changed and stop.
  4. Action each request in document order, with a surgical Edit to the passage it refers to. For an actioned @claude: comment, remove that comment in the same edit. Follow the Editing rules below.
    • If a request needs information not present in the sources, do not invent it. Leave any related comment in place (so it stays outstanding) and note the gap for the wrap-up — e.g. "the cited sources don't state X". Make only the part of the change the sources do support.
    • If a request is ambiguous, ask the user instead of guessing.
  5. Confirm completion. Report what you changed (one line each) and, separately, any gaps — requests left unactioned because the sources don't cover them — so the user can add a source or revise.
  6. Tutor re-test reminder. After confirming, check whether the edited page belongs to a learning course and, if so, respect the user's chosen cadence (see Tutor prompt trigger below). Skip entirely if no course covers the page.

Editing rules

  • Ground edits in the sources only. New or revised factual claims must be supported by the page's cited sources (step 2). Match the page's existing citation style (inline relative links to ../sources/... / ../notes/...). Pure wording/tone edits that add no new facts are always fine.
  • Preserve Obsidian/Markdown syntax exactly. Treat these as literal content — never reformat, normalise, or rewrite them unless a request explicitly asks:
    • YAML frontmatter (the --- block): keep every key, value, and ordering as-is unless asked to change a specific field.
    • [[wikilinks]], [[link|alias]], embeds ![[...]], block refs ^block-id, heading links [[note#heading]], #tags, and callouts (> [!note]).
    • LaTeX math ($ ... $, $$ ... $$) — leave existing math untouched unless a request targets it.
  • Write math in LaTeX. When a request adds or revises math, use LaTeX (inline $ ... $, display $$ ... $$, which render in Obsidian and VS Code) — never plain text or Unicode. If asked to make something more precise or rigorous, describe it mathematically and include the critical equations/derivations, grounded in the page's sources.
  • Only change what's asked. Edit the specific passage and nothing else. Leave other text, spacing, and line breaks untouched. Don't restructure headings or rewrap paragraphs unless asked.
  • Edit in place. Change the original file at $ARGUMENTS — never create a new file, add a suffix, or change the location.

Tutor prompt trigger — re-test reminder

After a successful edit, check whether the page is covered by a learning course and remind the user to re-test at their chosen cadence:

  1. Glob course/*/goal.md. For each, check whether the edited page's slug/title appears in that course's Knowledge Audit, or the page lives under the course's own modules/. If no course matches, do nothing.
  2. For a matching course <slug>, read its tutor_prompt_mode (from goal.md):

    • session-end — at the very end of the wrap-up (once per session, even if several pages in the course are edited), append:

      Course <slug> covers this page — remember to run /tutor <slug> this session. Track with a flag file course/<slug>/.session_reminded (treated as created on the first reminder this session); don't repeat the line if it already exists.

    • after-3-edits — read the integer in course/<slug>/.edit_count (0 if absent), increment, write it back. When it reaches 3, prompt

      You've edited 3 pages covered by course <slug> this session — run /tutor <slug>? and reset .edit_count to 0.

    • never or unset — no prompt.

These flag files are gitignored runtime state, not knowledge-base content.

Notes

  • Works from the page and its sources — no API key. The user iterates by adding more @claude: ... comments, chatting, or editing directly, and re-running the skill; each run actions the outstanding requests (leaving only those blocked by a source gap).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.