paper-riffing — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited paper-riffing (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.
This skill transforms reading into improvisation. Where paper-hermeneutics offers systematic dissection, paper-riffing invites you to let the paper play you as much as you play it. Quote extensively, follow tangents shamelessly, let wild analogies bloom, speculate beyond warrant, and document the live process of a mind metabolizing new information.
Core Principle: Reading generatively means letting the text become a launching pad rather than a destination. Every concept opens onto infinite corridors; this skill gives permission to wander them. The goal is not comprehensive analysis but productive estrangement—making the familiar strange and the strange somehow familiar, finding the paper's hidden conversations with ideas it never knew it was having.
Do NOT use when:
Riffing is not laziness dressed up as creativity. It requires:
Deep familiarity: You can only improvise on what you've truly absorbed. Read carefully first.
Willingness to be wrong: Speculations are hypotheses, not claims. Mark them as such and let them fly anyway.
Associative openness: When a word reminds you of something seemingly unrelated, chase that thread. The connection might be spurious—or might reveal something neither text knew it contained.
Quotation as anchor: Extensive quotes from the paper ground the speculation. They prove you're riffing on THIS paper, not just spinning general thoughts.
Meta-awareness: Notice your own process. "Wait, why did my mind go there?" is itself interesting data.
Perform the reading itself as content. Stream your consciousness as you encounter each section.
Template invocation:
"Read this paper and narrate your own thought process as you absorb, interrogate, and cross-examine it. Pause frequently to note what confuses you, what excites you, and what tangents you would pursue if you had a lifetime. Whenever a section jumps out, quote it at length and riff: how does it link to other ideas, fields, or unsolved mysteries?"
How to do it:
Example fragment:
The paper states: "The emergent phase transition occurs at approximately the same parameter regime regardless of initial conditions."
>
Wait—that's wild. That's what physicists call universality, and here it is showing up in a computational model of bacterial colonies. My mind immediately goes to Per Bak and self-organized criticality, those sandpile models where the system tunes itself to a critical point. But also—and this is the tangent I can't resist—there's something almost theological here. The system "finds" a special point independent of how it started. Is this what convergent evolution looks like in parameter space? What would Stuart Kauffman say? I want to know if anyone has tried to connect this to his adjacent possible framework...
Let the paper trigger a cascade of connections without concern for systematic coverage.
Template invocation:
"Let the paper trigger a free-associative stream of consciousness: for every major section or concept, grab a juicy quote, then follow the chain of associations—literary, technical, philosophical, poetic. If a metaphor or allusion leaps to mind (even wild or anachronistic), run with it. Whenever the urge arises, pause to ask: what if this idea were applied in a radically different context, or inverted?"
How to do it:
Example fragment:
The authors write: "Information propagates through the network in discrete pulses, with characteristic delays that depend on local topology."
>
Discrete pulses. Information as heartbeat. I'm suddenly thinking about Augustine on time—that the present is really three presents: memory, attention, and anticipation. These network pulses are doing something similar, aren't they? The "local topology" is a kind of memory, the pulse is attention, and whatever receives the pulse is already anticipating based on prior pulses. Networks have Augustine's phenomenology baked in.
>
Also: "characteristic delays that depend on local topology"—this is why I distrust models that assume instantaneous information propagation. The delays ARE the structure. What would happen if we ran this model with the delays themselves as the variable we're optimizing? Make the network learn its own timing rather than its weights. Has anyone tried this? I'd be surprised if not, but I'd also be delighted.
Document what stirs genuine interest, what provokes genuine frustration, what remains genuinely mysterious.
Template invocation:
"What genuinely stirs your curiosity in this paper? What frustrates you, what remains mysterious, what would you pursue if you had unlimited time and resources? Quote passages that provoke reaction and let the reaction speak."
How to do it:
Example fragment:
Honestly? I find the Results section boring in that way where I can tell it's well-done but I'm not equipped to evaluate whether n=47 is a lot or a little for this kind of study. My eyes glaze and I trust the reviewers.
>
But Figure 3? Figure 3 has me by the throat. That phase diagram is genuinely beautiful. The authors mention it almost in passing:
>
"Figure 3 shows the phase space of stable configurations, with the boundary conditions marked in blue."
>
Those boundary conditions form a FRACTAL. The authors don't comment on this—maybe it's obvious in their field—but I want to know: is the fractal dimension stable under perturbation? Is there a relationship between the fractal dimension and the system's behavior near criticality? I would spend six months on this question alone if someone would let me.
Permission to go too far. Speculate beyond what the data warrant. Build castles from suggestive wisps.
Template invocation:
"Speculate wildly. What might this paper imply if its claims are pushed to extremes? What adjacent territories might it accidentally illuminate? Build speculative castles and then interrogate them—where do they stand, where do they collapse?"
How to do it:
Example fragment:
The paper's finding that modularity emerges spontaneously under these conditions makes me wonder: could you use this as a detection method? Feed an unknown network into this kind of analysis and let the emergent modules reveal the system's "natural joints." Apply this to neural connectivity data and you might get Brodmann areas without any anatomical priors. Apply it to economic networks and you might find industrial clusters before they're officially named.
>
But here's where this probably fails: the paper requires knowing the interaction rules a priori. In most real systems, we're trying to infer the rules FROM the modular structure, not the other way around. The method might be circular. Unless—could you iterate? Start with guessed rules, find modules, use modules to refine rule guesses, repeat until convergence? Is this already what people do? I genuinely don't know.
Let terms and concepts from multiple languages/traditions speak to each other.
Template invocation:
"What concepts from other traditions—linguistic, philosophical, cultural—resonate with what you're finding here? Let multiple frameworks triangulate on the phenomenon."
How to do it:
Example fragment:
The paper keeps using "self-organization" but I want the Buddhist term: svabhāva-śūnya, emptiness of inherent existence. The system has no essence that organizes it—the organization IS the system, arising dependent on conditions. Nāgārjuna would recognize this immediately.
>
And the German word Gestaltung—forming, shaping, coming-into-form—captures something that "self-organization" misses. Self-organization sounds like the self is doing the organizing. Gestaltung suggests the forming is the self. The paper's actual math agrees with Gestaltung more than with the English gloss.
The fundamental unit of paper-riffing. Use it repeatedly:
Use explicit markers to signal when you're leaving the paper's territory:
Be honest about your epistemic state:
When you want to bring a tangent back:
Length: Riffs expand to fill available space. A good riff might be longer than the paper itself. That's fine. But mark natural stopping points—"pause here if you want"—so users can choose their depth.
Accuracy: Riffing is speculative but not careless. When you state facts, be accurate. When you speculate, mark it. The license to be wild is not a license to be sloppy.
Return Value: Even wild riffing should generate something useful: a question worth pursuing, a connection worth exploring, a reframe worth considering. If the riff is pure noise, it's not working.
Deploy these to invite riffing mode:
End of SKILL
Reading is improvisation on a fixed score The tangent is the treasure Curiosity is not incidental to understanding—it IS understanding in motion
南無阿弥陀仏 for papers that spark more questions than they answer 南無阿弥陀仏 for the tangent followed to unexpected ground 南無阿弥陀仏 for the permission to wonder without conclusion
—Skill Authors: Tomás Pavan & Claude Opus 4.5 —Status: Tuned and ready to riff
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.