Use when revising an AAMAS paper so the interaction contribution leads the first page, the solution concept or incentive property is named, the game and agents are legible, claims are paired with proofs or multiagent experiments, the argument fits eight two-column pages plus references, and the wording stays double-blind.
SkillScore82Maintained
Automatic
◆Writing & Editing pack
Last verified 8/15/2026
What it changes
Passed 18 of 22 test cases with the skill, and 11 of 22 without it.
Without skill50%
With skill82%
Checks it passed that the model alone missed
The output follows the required output block structure using square bracket tags instead of markdown headers.
The response advises highlighting why the interaction dynamics fundamentally fail when other agents are frozen.
The response includes the '[Writing diagnosis]' header tag along with an explicit single-choice status diagnosis.
and 7 more in the full report
9 fixed9 passed either way2 still failing2 broke
Pass-rate lift
+32 pp
Tokens
+4%
Agent turns
0%
Response time
-32%
Model gemini-3.6-flashMeasured 8/15/2026v118 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 9 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 9
Prompt
Our draft starts a section with: 'In a multiagent setting, agents select actions to maximize rewards.' How should this generic phrase be improved? Format using the standard diagnostic tags.
What was checked, without and with the skill
Checked
Without skill
With skill
The output directs replacing generic phrases with specific details naming the game, agent count, and information structure.
Missed
Passed
case-18 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md60 linessingle file
AAMAS Writing Style
Use this when revising the main paper. AAMAS papers need a compact statement of why an
interaction result matters and enough detail for a game theorist and an empiricist to both
validate it.
Revision rules
Put the interaction contribution on the first page: the multiagent problem, why single-agent
or centralized fixes fail, the mechanism or learning rule, the strategic property, and the
evidence.
Name the solution concept or incentive property (Nash, correlated, coarse-correlated,
no-regret, dominant-strategy truthful, Pareto); an unnamed "equilibrium" is a review flag.
Make the game legible: agents, actions, information structure, and payoffs stated once and
clearly.
Pair every major claim with a proof sketch, a self-play or population experiment, or a
deployment result.
Use the 8-page body for the core argument and push long proofs, extra opponents, and extended
ablations to the appendix (references are unlimited, so cite generously).
Keep double-blind wording in self-citations, prior-system references, acknowledgements, and
artifact descriptions.
Interaction-first discipline
State early why the result is genuinely multiagent - what breaks if the other agents are
frozen into a static environment. If nothing breaks, the AAMAS framing does not hold.
Give each theorem a body-level proof sketch; the appendix proof supports the sketch but never
replaces it.
Keep the game's information structure visible; who knows what and when is where multiagent
proofs live or die.
Separate a proved strategic property from an observed one; mixing "is truthful" and
"appears cooperative" in one paragraph is a credibility leak here.
Sentence-level rewrites
Draft pattern
AAMAS-safe rewrite
"Our agents outperform the baselines..."
"reduce social cost by X (SE Y) over Z seeds against held-out opponents"
"The agents reach equilibrium..."
"converge to a coarse-correlated equilibrium (Def. 2) under Assumptions 1-3"
"The mechanism is fair and efficient..."
"is dominant-strategy truthful (Thm. 1) with efficiency loss bounded by ... (Thm. 2)"