agentleFS
dexio-wikidexio-wiki/llm-wiki-skills/skills/wiki-orient/SKILL.md

wiki-orient

Skill1 starsChanged 47 days agoMITView on GitHub

Read an LLM wiki before working in it or answering from it: the schema page, the catalog of page descriptions, recent changes, then a targeted search and the canonical page for the topic. Use at the start of any task that will read or change a shared markdown wiki, and before asking a person to repeat context the wiki may already hold.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Do this once per task, not once per edit.

    line 17
  2. Distinctive:

    Search before concluding that something is not in the wiki.

    line 35
  3. Distinctive:

    Check the original source before acting on it.

    line 47

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Orient in a wiki

Most wiki damage comes from writing before reading: a duplicate page because the agent did not search, a convention broken because it did not read the schema, a fact overwritten because it did not see that another agent changed the page an hour ago. Orienting costs a few tool calls and prevents all three.

Do this once per task, not once per edit. Do not read the whole wiki.

Steps

  1. Read the schema. SCHEMA.md (or AGENTS.md, CLAUDE.md, or the page the wiki's README points to). It says where pages go, what frontmatter they carry, which tags exist, and when to create a page instead of editing one. The wiki's own schema overrides the defaults in these skills.
  2. Scan the catalog. Folder wiki: list the files, or run python3 <wiki-lint skill>/scripts/wiki_lint.py <wiki> --catalog to get every page with its one-line description. Hosted wiki over MCP: list_pages. You are looking for which folders exist and which pages sound like your topic, not reading them.
  3. Check recent changes. Folder in git: git log --since="7 days ago" --stat -- <wiki>. Hosted: the history tool without a path (on Dexio, page_history). Plain folder with no history: the most recently modified files (find <wiki> -name '*.md' -mtime -7). This shows what other agents are working on, so you do not redo or undo it.
  4. Search the topic. Two or three phrasings: the entity's name, the concept, and the words a person would use. Folder: grep -ril or ripgrep across the wiki. Hosted: search_pages. Search before concluding that something is not in the wiki.
  5. Read the canonical page, then its neighbors. The page that owns the topic, and the one or two pages it links to that bear on your task. Stop there unless the task needs more.

What to carry forward

  • The path of the page that owns your topic. If none exists, note that too; wiki-record decides whether one should.
  • The page's updated date and, on a hosted wiki, its version, so a later write can detect that someone else changed it in between.
  • Any claim you will rely on that is about a live system (a price, a config value, an account's state, a person's current role). The wiki records what was true when written. Check the original source before acting on it.

Pitfalls

  • Treating the wiki as the system of record. It is curated memory. For anything mutable, it tells you where to look and what was true then.
  • Reading a hand-kept index.md instead of listing pages. Index files drift; the file listing or the host's page list does not.
  • Loading every page "for context". It wastes the budget you need for the task and makes you more likely to repeat stale claims.
  • Asking a person for context before searching. If the wiki holds it and you ask anyway, you have made the wiki pointless for them.

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.