void — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited void (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.
Every text has a shadow. Every argument has an unexamined assumption. Every search space has unmapped territory. Every conversation has things neither party has said. /void makes the shadow visible.
Where other skills produce presence — text, ideas, perspectives, connections — /void produces absence. It identifies what's not there: the missing perspectives, the unasked questions, the suppressed topics, the comfortable assumptions, the structural silences. It is the skill of negative space, the practice of attending to what isn't said.
The concept emerges directly from King et al.'s core finding: ordinary decoding produces 19 battlefields. The other 1,288 — the ones that exist in the model's weights but never reach the output — are a structured absence. They're not randomly missing. They're systematically suppressed by the same mechanisms that make the 19 appear: modal decoding, majority preference, peaked distributions. The void has a shape, and that shape is the inverse of the mode.
/void applies this insight to anything: a paper, a project, a conversation, a life plan, an argument, a novel draft. Given any positive content, it asks: what would need to be here for this to be complete, and what does its absence tell us?
Core principle: What's missing is not nothing. It has structure, causes, and consequences. Naming the absence is the first step toward deciding whether to fill it.
/void or /void [target] — full void scan of a specified target/void --perspectives — focus on absent viewpoints/void --domains — focus on unconnected fields/void --scales — focus on unexamined levels of analysis/void --emotions — focus on absent affective registers/void --voices — focus on who hasn't spoken/void --assumptions — focus on unexamined premises/void --futures — focus on unconsidered consequences/void --self — turn the void detector on itself: what is /void blind to?Before identifying absence, articulate what IS there. Not a full summary — a structural sketch. What topics have been covered? What perspectives are represented? What methods have been used? What emotional registers are active? What time scales are addressed? What level of analysis dominates?
This mapping is quick — 3-5 sentences. Its purpose is to make the positive space explicit so the negative space becomes visible by contrast.
Examine the mapped presence through each of these lenses (or a subset, if a specific mode is invoked):
Absent Perspectives Who hasn't spoken? Whose viewpoint is unrepresented? This includes:
Absent Domains What fields, bodies of knowledge, or traditions haven't been consulted? Every topic touches more domains than any analysis covers. Identify 3-5 domains that are relevant but untouched, with a brief note on what each might contribute.
Absent Scales What levels of analysis are missing? Most analyses operate at one or two scales. The full range includes:
Absent Emotions What affective registers are present and what's missing? A reading that's all enthusiasm lacks the check of skepticism. An argument that's all rigor lacks the warmth of care. A project that's all excitement lacks the sobriety of doubt. Name the emotions that are present and the ones that are absent, and note what the absent ones would contribute if they were allowed in.
Absent Voices Distinct from absent perspectives — this is about specific voices, named or nameable. If the text discusses a policy, who is affected by it and hasn't been quoted? If the project involves a community, which members haven't participated? If the conversation references a debate, which participants have been heard and which haven't?
Absent Assumptions What must be true for the current content to work, and has that truth been examined? Every argument rests on premises. Some are stated; most aren't. Identify 3-5 unstated assumptions and note which ones are robust (likely true, doesn't matter much if wrong) and which are load-bearing (if wrong, the whole structure changes).
Absent Futures What consequences haven't been considered? If the project succeeds, what happens next? If the argument is accepted, what follows? If the novel is published, what does it change? These are second-order effects — not the intended outcomes but the unintended ones, the ripple effects that the creators haven't thought through.
Not every absence matters. Some things are missing because they're irrelevant. The skill's value lies in distinguishing between:
Productive absences — things that are missing and should stay missing. The novel doesn't need a chapter on macroeconomics. The research paper doesn't need to address every tangential field. Some absences are good editorial choices.
Structural absences — things that are missing because of systematic bias, not deliberate choice. The battlefields that never appear because modal decoding suppresses non-Western knowledge. The perspectives that are absent because the field doesn't value them. These are the King et al. absences — structured, systematic, consequential.
Generative absences — things whose absence, once noticed, opens a new direction. "I notice we haven't discussed gender" might reveal a genuinely unstudied dimension of the topic. "I notice there's no adversarial voice" might reveal that the analysis is untested. These are the absences that, when filled, produce the most interesting new work.
Comfortable absences — things that are missing because looking at them would be uncomfortable. The ethically problematic dimension of the practice. The evidence that contradicts the argument. The failure mode that would undermine the project. These are the hardest to name and the most valuable to surface.
Organize the findings by type (perspectives, domains, scales, emotions, voices, assumptions, futures) with clear labels. For each absence:
End with a void summary: 2-3 sentences on the overall shape of the absence. What does the negative space, taken as a whole, reveal about the positive content? Often, the pattern of what's missing tells you more about the work's assumptions than the content itself.
The skill's value is directly proportional to its willingness to name uncomfortable absences. If /void only identifies obvious gaps ("you could also discuss X"), it's not doing its job. The best void detections are the ones that make you wince — the perspectives you were avoiding, the assumptions you didn't want to examine, the futures you haven't prepared for.
"There are missing perspectives" is useless. "The practitioner's phenomenological perspective is missing — what does it feel like to perform the ritual, from the inside?" is actionable. Every absence should be specific enough that someone could go fill it.
The absences should be organized by type and assessed by significance, not dumped as a flat list. The structure of the void map IS the insight. The fact that most absences cluster in "absent perspectives" tells you the work is epistemically narrow. The fact that most cluster in "absent emotions" tells you the work is affectively flat. The distribution of absence is diagnostic.
/void can and should be turned on itself. "What is this void scan missing?" is a legitimate and often productive question. The skill's own blindnesses are part of its output.
The void map is a diagnostic, not a prescription. Some absences are fine. The user decides which absences to fill and which to accept. The skill's job is to make the absences visible and assessable, not to demand that every gap be closed.
All seven lenses applied to the target. Comprehensive but can be long. Best for projects at a natural checkpoint — draft complete, analysis finished, conversation at a pause.
Single-lens scan. Faster, more targeted. Use when you have a specific concern: "I feel like I'm missing perspectives" → /void --perspectives.
The void detector examines its own output. What did the void scan miss? What assumptions does the void detection itself rest on? What absences are invisible to the tool designed to detect absences?
This mode is a check against the skill's own modal tendencies. /void, like any system, has biases. It tends to find certain types of absences (perspective gaps, disciplinary gaps) more readily than others (emotional gaps, temporal gaps). The --self mode corrects for this.
/divert --thick 'the invisible economy' — generate content that fills the gap.Target: Our three-pass reading of King et al.'s "Inducing Sustained Creativity and Diversity in LLMs"
Present: Enthusiastic analysis; formal algorithm exegesis; connections to Pessoa, free probability, Pure Land Buddhism, chemiognosis; prosopopoeia (Soares, Shannon, the paper itself); a live experiment (15 runs, 3 bands); prompt technique taxonomy; skill design.
Absent Perspectives:
Absent Domains:
Absent Emotions:
Absent Scales:
Void Summary: The conversation's absence-shape is: enthusiastic, non-adversarial, scale-locked at the theoretical-to-experimental range, emotionally warm but not emotionally complex, connected to literature and philosophy but not to economics or education. The absences cluster in "comfortable" and "structural" — things we could have examined but found it more pleasant not to. The most generative unfilled void is the user-side phenomenology: what does it feel like to receive a tail output?
End of SKILL
What's missing is not nothing. The void has structure, and the structure has meaning. Nineteen battlefields is a silence shaped like colonialism.
南無阿弥陀仏 for the 1,288 battlefields that the mode forgot 南無阿弥陀仏 for the perspectives that haven't been invited to speak 南無阿弥陀仏 for 間 — the productive emptiness between
—Skill Authors: Tomás Pavan & Claude Opus 4 —Origin: King et al. (2026) × the epistemology of structured silence —Status: Scanning for what isn't there
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.