Assumption Surface

Assumption Surface

A focused, atomic skill: extract the hidden assumptions that an external text takes for granted. Where argument-deconstruct builds a full Toulmin layout, this skill produces a flat assumption table — faster, broader, less structural.

The output is a working surface for the user to inspect, not a finished argument. It is what you'd run before a board meeting to prepare counter-questions, or before a contract signing to spot what the other side is quietly assuming you'll grant.

When to use

Trigger phrases (any language):

• 「揭露這份備忘錄的隱性假設」「這個策略在假設什麼」「stress-test 這些主張」
• "find the hidden assumptions", "what is this assuming", "stress-test these claims before deciding"
• "surface the implicit world-model"

Skip when:

• User wants a full argument deconstruction with claim-grounds-warrant chain — use argument-deconstruct
• User wants all design layers (rhetoric / frame / persuasion / etc) — use artifact-deconstruct
• User wants to doubt their own premises — use philosophers-toolkit:descartes-methodical-doubt
• Target is

Source claims (numbered)

1. ...
2. ...
3. ...

...

Assumption table

| # | Assumption | Source claim(s) | Strength | Falsifiability test | |---|---|---|---|---| | 1 | Assumes that ... | claim 1 | Foundational | Falsified if ... | | 2 | Assumes that ... | claim 2 | Load-bearing | (n/a if not foundational) | | 3 | Assumes that ... | symptomatic reading | Decorative | (n/a) | | ... | ... | ... | ... | ... |

Foundational assumptions worth challenging

For each foundational assumption, propose a counter-question:

• Assumption 1: . Counter-question: ...
• Assumption N: ...

Bottom line

## Lenses available

This skill ships with 1 self-contained lens reference:

- [`references/lens-symptomatic-reading.md`](references/lens-symptomatic-reading.md) — Althusser-influenced "reading the absent" (1968 / 1970 *Reading Capital*)

Toulmin appears in this skill via the reverse-Toulmin protocol (which
inverts the standard direction). The full Toulmin reference lives in
`argument-deconstruct/references/lens-toulmin.md`. Per ADR-0002,
each skill is self-contained — this skill's protocol embeds the
relevant Toulmin material rather than cross-referencing.

## Sample fixtures

- [`assets/sample-company-strategy-memo.md`](assets/sample-company-strategy-memo.md)
- [`assets/sample-tweet-thread-productivity.md`](assets/sample-tweet-thread-productivity.md)

## Anti-patterns

- **Confusing "claim" with "assumption"**: a claim is what's *said*; an assumption is what's *taken for granted to make the claim work*.
- **Inflating assumption count**: more is not better. 5–15 is the sweet spot. 30+ assumptions usually means you're listing claims as assumptions.
- **Decorative-only output**: if all your surfaced assumptions are "decorative," you have not pressed hard enough — every text rests on at least one foundational assumption.
- **Skipping falsifiability test**: foundational assumptions without falsifiability tests are not yet useful for the user. They are just labels.
- **Treating symptomatic reading as gotcha-hunting**: the goal is to surface what's structurally absent, not to catch the writer in tonal slips. Stay structural.

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [kouko](https://github.com/kouko)
- **Source:** [kouko/monkey-skills](https://github.com/kouko/monkey-skills)
- **License:** MIT

Install and usage instructions live in the source repository linked above.