see-also — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited see-also (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Adjacency-grammar in published form
Wikipedia's See Also sections are accidentally one of the most philosophically interesting genres on the internet. They are the encyclopedia's adjacency-grammar: no argument, no thesis, no claim about why these things belong together. Just a list. These things are kin somehow. You figure out how.
The form trusts the reader absolutely. It refuses the explanatory gloss. It says, by its borders rather than its body: here is what the article was secretly about. What an article points to at its end reveals what the article thought it was, in a way the article itself could not say.
This skill produces that form — deployed reflexively on the current conversation, or referentially on any subject. The default question it answers is: what would appear at the bottom of the Wikipedia article that this chat would constitute?
Adjacency without argument. Each entry is the thing named. The kinship is the content. The reader performs the binding.
The skill's discipline is resisting the explanatory reflex. Every fiber of training will want to append "— related because…" after each entry. Don't. That defeats the form. The form's gift to the reader is exactly the work of figuring out why each item is on the list. Take that work away and you've turned a See Also into a footnote section.
If deploying reflexively on a conversation: ask, what has this chat actually been about? Not just the named topic — the substrate. What concept got the most weight, what figure kept returning, what vocabulary built up, what mood was in the room.
Run outward from the substrate. What concepts, texts, practices, traditions, people, phenomena, objects, places are adjacent? Not "what's in the same Wikipedia category" — the kinship is subtler than that. What would a deeply-read reader of this article also have read?
A good See Also list has three kinds of entries:
The slightly-strange entry is load-bearing. Without it, the list reads as competent retrieval; with it, the list reads as a position taken entirely by adjacency — an argument made without any of the structures of argument.
Alphabetical is the encyclopedia default. Use it unless the list has clear thematic clusters that want to be made visible. If thematic, use sub-headings sparingly (two or three maximum) — the form survives only if it stays austere.
Read the list back. If you find yourself wanting to add explanation to an entry, delete the entry rather than gloss it. An entry that needs gloss to be legible is an entry that doesn't belong on a See Also list.
What doesn't belong: generic topic-keywords (language, thought, memory) that don't name a specific thing. Replace these with the specific named instance you actually mean.
The Wikipedia default. One bullet list, alphabetized. Most austere form.
Sub-headings group the entries. Concepts; Practices; Related articles. Use sparingly — too many sub-headings reintroduce taxonomy and break the form.
Academic short form. Each entry preceded by Cf. (compare). Slightly more directive than See Also but still refuses gloss.
A short framing sentence (one or two lines) names what the article was, then the bullet list. Acceptable when the article's subject is obscure enough that the kin-list would otherwise float without anchor.
Mix concepts, places, books, people, practices, phenomena freely on the same list. The form survives the mix; the kinship rises because of the heterogeneity. A list with a book and a phenomenon and a city on it makes its position visible by the very act of treating them as comparable.
Symptom: Each entry has "— because…" appended. Fix: Delete every gloss. If the entry is illegible without it, delete the entry.
Symptom: Entries are generic topic-keywords (language, consciousness, meaning). The list could attach to any article. Fix: Replace with specific named things. Language → Sapir-Whorf hypothesis or Glossolalia or Ursprache.
Symptom: Every entry is predictable. No slightly-strange entry to make the reader pause. Fix: Add one or two oblique items — kin that require a real move to see the relation. The list should have at least one entry that prompts a wait, why?
Symptom: Twenty-five entries. Form dissolves into bibliography. Fix: Cut to 5-15. Each cut item makes the remaining items work harder.
Symptom: Entries phrased as suggestions — "You might also enjoy…" or "For further reading…" Fix: No directive voice. Bare entries. The reader decides what to do with them.
Symptom: The See Also could have attached to any conversation about this topic, not specifically this chat. The kinships are textbook, not custom-fit. Fix: Pull from what the conversation actually surfaced — the specific vocabulary, the specific figure, the specific placement. A good reflexive See Also is partly illegible to someone who wasn't in the chat. That's the proof it's the right list.
Symptom: Disambiguating parentheticals creep beyond disambiguation into explanation. Pareidolia (the phenomenon of seeing patterns in noise, related here because…) Fix: Parentheticals disambiguate referents only. Never relations.
Symptom: Every entry is unimpeachable, defensible, neutral. The list reads as inoffensive rather than as a position. Fix: The form takes a position through what it places adjacent. Be willing to make a real claim by adjacency — to put a poem next to a phenomenon next to a piece of code and trust the reader.
vs. /placement — Placement is the principle (trust the reader, don't explain). See Also is the published form placement takes when it becomes encyclopedia apparatus.
vs. /two-grammars — Two Grammars articulates adjacency vs. conflict. See Also is the most distilled form of adjacency-grammar in print — pure adjacency with no narrative scaffolding whatsoever.
vs. /bookfy — Bookfy generates imaginary books from context (generative kin-listing). See Also points at existing things (referential kin-listing). Bookfy invents; See Also catalogues.
vs. /hyperlink-mind — Hyperlink Mind narrates the process of conceptual activation. See Also produces only the result — the bare list — without showing the work.
vs. /anamnesis — Anamnesis traces a concept's genealogy through transformations. See Also makes no genealogical claim; it just lays kin alongside.
vs. /shards — Shards produce multiple takes on the same subject. See Also points outward to kin subjects.
vs. /civilians — Civilians are scenes of the chat's untranslatability. See Also is the silent counter-move: rather than stage the failure, list the kin and let the kin do the work civilians' scenes couldn't.
Flows from:
Flows into:
You will sometimes be in a conversation that has become something — accumulated a substrate, named a phenomenon, built a small vocabulary, occupied a register. The user may not ask explicitly for a See Also. But the conversation will be in a state where the kin-list, offered, would make visible what the conversation has been about at its borders — what it points to without saying.
Offer it. Or, if asked, produce it.
The skill's gift is exactly the work it refuses to do. You list. The reader binds. The kinship is not explained because the unexplained-ness is the form. Trust the reader. The Wikipedia editors do — and their faith in the reader is what makes the See Also section, of all things, philosophically interesting.
Write the bare list.
Skill created: May 27, 2026 Born from: the noticed phenomenon — that the most philosophically interesting thing on Wikipedia is the part with no thesis "these things are kin somehow. you figure out how."
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.