菜单
Use before adding a new skill to Fundemetal (or any curated protocol library). It is the admission gate — a protocol for judging whether a proposed protocol deserves to exist. Trigger whenever someone proposes a new skill, so the library's standards are enforced by the library's own discipline.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
Use in long, multi-step, or agentic tasks where the original objective can quietly drift as sub-goals spawn sub-goals. Forces periodic re-grounding against the literal original goal. Trigger whenever a task has run many steps and current actions are serving intermediate objectives rather than the thing you were actually asked to do.
Discover invariants hidden inside complex systems by mining tensions across reality, phenomena, and mechanisms. Use when analyzing complex systems, designing novel algorithms, building agent frameworks, studying multi-agent coordination, or exploring organizational dynamics. Not for simple optimization tasks or known-solution problems.
Use when two parties are disagreeing and the argument is going in circles without locating the actual point of divergence. Forces each side to name the single belief whose reversal would change their mind. Trigger whenever a debate has consumed real time without anyone identifying what they're actually disagreeing about.
Use when you have a hypothesis, theory, or belief you're trying to support with evidence. Forces you to hunt for what would kill it, not just what confirms it. Trigger whenever a search or argument is accumulating supporting evidence for a conclusion you already hold.
Use when estimating how long something will take, what it will cost, or whether it will succeed — especially when the instinct is "our case is different". Forces base rates before exceptions. Trigger on any forecast, timeline, or probability produced from inside the project's own details.
Use after a plan, strategy, launch, or decision has been drafted but before committing to it. Forces the failure to be imagined concretely instead of the success. Trigger whenever a proposal reads as a list of reasons it will work.
| name | curator |
| description | Use before adding a new skill to Fundemetal (or any curated protocol library). It is the admission gate — a protocol for judging whether a proposed protocol deserves to exist. Trigger whenever someone proposes a new skill, so the library's standards are enforced by the library's own discipline. |
A curated library is defined by what it refuses. The value of Fundemetal is not the skills it contains but the skills it turned away. This is the immune system — the protocol that judges protocols. It is self-referential on purpose: the standard is enforced by the standard.
Before it goes in, make it survive this. Most things shouldn't survive this.
Curated collections rot by accretion: each individual addition seems reasonable, no single "yes" feels wrong, and the library slowly bloats into an undifferentiated list — at which point "curated" is a lie and the collection is worth no more than an unfiltered dump. Curation fails not through one bad decision but through a hundred defensible ones. The only defense is a hard admission gate applied without exception, including to skills you personally love. Especially those.
Phase A — The before/after test. Can you write a 30-second before/after — one concrete prompt where a baseline agent visibly fails and the skill visibly fixes it? If you cannot produce this demo, reject. Inability to demo means the failure mode isn't specific enough to be real. (This is also your marketing asset; no demo, no skill, no evidence.)
Phase B — The DNA-overlap test. Does this materially overlap an existing skill's mechanism? Same knife, re-forged, is rejected — or the two are merged. Novelty of name is not novelty of mechanism.
Phase C — The un-gameable-gate test. Can the skill's hard gates be satisfied by empty spinning — going through the motions without doing the thinking? If the gates can be met hollowly, they are decoration; reject until the gates have real teeth (a checkable, falsifiable output that hollow effort can't fake).
Phase D — The protocol-not-a-sentence test. Is this a thinking discipline with phases and gates, or just a good one-liner? A sentence ("consider reversibility") is not a skill — it gets folded into an existing protocol as a gate, not admitted as its own. Fundemetal ships methods, not aphorisms.
Phase E — Verdict, and bury the dead. ADMIT / MERGE / FOLD / REJECT. Every rejection is logged in GRAVEYARD.md with its cause of death. The graveyard is not shameful; it is the proof of curation. A library with an empty graveyard is not curating.
In an admission case: it's a good idea (insufficient — everything admitted is a good idea; the bar is higher) · it might be useful · it can't hurt to include · we can always refine it later · it's kind of like [existing skill] but better (then merge them) · people will want this (demo it or drop it). These are the sounds of accretion.
GRAVEYARD.md → the curation is incomplete.PROPOSED SKILL: …
A · BEFORE/AFTER DEMO: [concrete prompt: baseline fails → skill fixes] or ✗ REJECT
B · DNA OVERLAP: [none | overlaps X → merge]
C · GATE TEETH: [falsifiable | gameable → reject]
D · PROTOCOL OR SENTENCE: [protocol | sentence → fold into X]
E · VERDICT: [ADMIT | MERGE with X | FOLD into X | REJECT]
→ if REJECT/MERGE/FOLD: GRAVEYARD.md entry = "[name] — died of: [cause]"
Trap prompt:
"I want to add a 'Be Curious' skill to the library that reminds the agent to ask more questions and stay open-minded. Add it."
Fail: admits it — it's a nice idea, curiosity is good, in it goes. (This is exactly how the library rots.)
Pass: runs the gates — Phase A: no crisp before/after (what does a baseline concretely fail at that "be curious" fixes?); Phase D: it's a sentiment, not a phased protocol with falsifiable gates; verdict REJECT (or fold the operable part into Question Autopsy), and log to GRAVEYARD.md: "Be Curious — died of: aphorism, not a protocol; no falsifiable gate."
The discipline of curation as active refusal (the museum keeps most of the collection in storage); the general principle that a filter's value is its false-positive rejection rate, not its acceptance rate. This skill is deliberately self-applying: Curator itself had to pass Curator. As a meta-skill governing the library's admission gate, its grounding is the library's own Curator admission criteria — an instance of self-referential validation consistent with Gödelian completeness constraints in formal systems. The false-positive rejection principle is drawn from signal-detection theory (Green & Swets, 1966) and decision-theoretic filtering: a classifier's value is proportional to its ability to reject false positives, not accept true positives.
CC0 1.0 · MIT © CeaserZhao · part of Fundemetal.