devils-advocate
/devils-advocate
Finds each conclusion sentence, classifies it (causal, numeric, absolute, forecast, recommendation) and writes a counter-argument from a template filled with that claim's own cause, number or subject, with an exposure score. Use this whenever the user states a conclusion and wants it argued against before acting. Pairs with steelman, skeptic-eye, swap-cause-effect and pov-core.
- Author
- max0wellseltzer-wow ↗
- Repository
- maestro-companions ↗
- Licence
- MIT
- Last repository change
- 26.10.2026
Assessment and security
- Files in this package
- 2
- Scripts
- 1
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 26.10.2026, without references to retired models or APIs.
MIT, open: use and modify while retaining author attribution.
- 1 scripts in this package. Review them before installing, as Claude may execute them.
Scripts: devils_advocate.py
What it does and when to use it
Finds each conclusion sentence, classifies it (causal, numeric, absolute, forecast, recommendation) and writes a counter-argument from a template filled with that claim's own cause, number or subject, with an exposure score. Use this whenever the user states a conclusion and wants it argued against before acting. Pairs with steelman, skeptic-eye, swap-cause-effect and pov-core.
Original author description
Finds each conclusion sentence, classifies it (causal, numeric, absolute, forecast, recommendation) and writes a counter-argument from a template filled with that claim's own cause, number or subject, with an exposure score. Use this whenever the user states a conclusion and wants it argued against before acting. Pairs with steelman, skeptic-eye, swap-cause-effect and pov-core.
How it works
- The tool (ships with this skill)
- Uses and limits
SKILL.md preview
---
name: devils-advocate
description: Finds each conclusion sentence, classifies it (causal, numeric, absolute, forecast, recommendation) and writes a counter-argument from a template filled with that claim's own cause, number or subject, with an exposure score. Use this whenever the user states a conclusion and wants it argued against before acting. Pairs with steelman, skeptic-eye, swap-cause-effect and pov-core.
---
# devils-advocate
A VANTAGE lens of the point-of-view family (engine: pov-core). It takes the other side of every conclusion. Each sentence with a verdict marker is classified by its most attackable form (causal, absolute, recommendation, forecast, comparison or numeric). It is negated into a COUNTER-CLAIM ("will" -> "will not", "always" -> "not always", "better" -> "worse") and given a counter-reason built from its own cause, effect or number. An exposure score out of 5 (absolute word, no number, no source, single cause, future tense) orders which conclusion to attack first.
## The tool (ships with this skill)
```
python devils_advocate.py --selftest
python devils_advocate.py <file|text> [--json]
```
```python
from devils_advocate import apply, negate
v = apply(note); v["findings"][0] # the most exposed conclusion and its counter
negate("We will launch next week.") # -> 'We will not launch next week.'
```
Example: `[devils-advocate] (text) 5 conclusions argued against; attack s1 first (causal, exposure 3/5)`, whose top finding is `[causal, exposure 3/5] s1: COUNTER We will not obviously launch the new screener next week because ... - 'we will obviously launch the new screener next wee' could happen without 'the bac...`
## Uses and limits
- The counters are arguments to test, not findings. A counter can be weaker than the claim.
- Negation is a word-level rewrite and can produce awkward grammar.
- Reads text and plans. Sentences without a verdict marker are left alone.
Files in this package (2)
- SKILL.md 2 KB
- devils_advocate.py 6 KB
Versions
| Repository | Score | Licence |
|---|---|---|
| This page · max0wellseltzer-wow/maestro-companions | 100 | MIT |
| daymade/claude-code-skills | 100 | MIT |
| griffinwork40/agent-framework | 100 | Apache-2.0 |
| mathews-tom/armory | 100 | MIT |
| Leuenaar/async-contractor-agents | 100 | MIT |
| Sellyew/Workbuddy-workflow | 100 | MIT |
| OneWave-AI/claude-skills | 100 | MIT |
| pedrohcgs/claude-code-my-workflow | 100 | MIT |
| cofin/flow | 100 | Apache-2.0 |
Database snapshot: 13.11.2026 · package version f86eec38e5db (SHA-256 digest of the entire package)
Downloads in the last 30 days: 1 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 100steelman
Extracts the claims of an opposing text, scores each for strength (number, mechanism, source, specificity), strips strawman markers and overreach and rebuilds the strongest version from the top claims. Use this whenever the user is about to dismiss a view he disagrees with.… - 100inner-critic
Writes the harshest specific critique question for each claim in a text, aimed at its weakest spot such as an unsourced number, an absolute word or a causal leap. Use this whenever the user drafts a claim or result note and wants the most cutting objection to each line. Pairs… - 90pm-devils-advocate
Structured challenge of product decisions with pre-mortem and counter-arguments. Use when someone says "challenge this", "devil's advocate", "poke holes", "stress test", "pre-mortem", "what could go wrong", "am I missing something", "sanity check this decision". Also use for…