lyndonkl/skills/propose-counterfactual
skills/propose-counterfactual/SKILL.md
--- name: propose-counterfactual description: Produces the counterfactual framing in an Intuition Builder 5-set — "what if this component were not here?" Reveals the function of a technical element by subtracting it and observing what breaks. Uses Pearl's causal ladder (counterfactual = level 3) as the theoretical spine. Use as the 5th archetype slot of generate-analogy-set, or invoked standalone when the writer wants to build intuition for why a specific element exists. Trigger keywords: counte
$ install --global
skillsauthnpx skillsauth add lyndonkl/claude skills/propose-counterfactualInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 25, 2026, 10:19 AM156.9s1 file scanned
SKILL.md
- name:
- propose-counterfactual
- description:
- Produces the counterfactual framing in an Intuition Builder 5-set — "what if this component were not here?" Reveals the function of a technical element by subtracting it and observing what breaks. Uses Pearl's causal ladder (counterfactual = level 3) as the theoretical spine. Use as the 5th archetype slot of generate-analogy-set, or invoked standalone when the writer wants to build intuition for why a specific element exists. Trigger keywords: counterfactual, what if not, remove, subtract, reveal function, why does this exist.
Propose Counterfactual
Table of Contents
- Workflow
- The 3 counterfactual moves
- Worked example
- Guardrails
Related skills: One of the 5 archetypes in generate-analogy-set. Can also be invoked standalone when the writer specifically wants the counterfactual angle without the full 5-set.
Workflow
For topic T:
- [ ] Step 1: Identify the component / mechanism whose function the writer wants to illuminate
- [ ] Step 2: Propose the subtraction — what if this component were absent?
- [ ] Step 3: Describe the concrete system that results (what you'd have instead)
- [ ] Step 4: Describe what breaks — performance, correctness, expressivity, efficiency
- [ ] Step 5: Return the counterfactual framing statement
The 3 counterfactual moves
Three sub-archetypes, pick whichever fits best:
- Ablation: "Remove X. What remains?" Reveals X's function by showing the degraded system.
- Substitution: "Replace X with its naive alternative. How does the system fail?" Reveals why X is the specific choice.
- Inversion: "What if X operated the opposite way?" Reveals directionality — why X pushes one way and not the other.
Worked example
Topic: Attention (in Transformers).
Ablation: "Remove attention from a transformer and you have a stack of residual MLPs per token — no information ever flows between token positions within a layer. The model can still transform each token independently, but 'context' is gone. That absence is what attention is 'doing.'"
Substitution: "Replace attention with fixed convolutions (the RNN/CNN alternative). You get locality — each token sees its neighbors — but the model can't arbitrarily connect token 1 to token 500. Attention's gift is not the operation, it's the arbitrary-range addressing."
Inversion: "Invert attention: instead of softmax-weighted averaging, what if the model picked exactly one token to copy from? That's hard-attention, and it turns out to be worse for training — soft interpolation makes the loss surface navigable. Attention is soft by gradient-descent necessity, not by design choice."
Pick one (usually ablation for a 5-framing set). The others can become a standalone post later.
Guardrails
- The counterfactual must be concrete. Not "imagine attention didn't exist" — actually describe what the system looks like without it.
- Describe the failure specifically. "Performance would suffer" is vague; "context is gone — the model processes tokens independently within a layer" is specific.
- Counterfactual framing is ≤2 sentences for the framing statement. The details can be expanded in the writer's draft.
- Do not confuse counterfactual with hypothetical. Counterfactual is "what if this weren't here" (revealing function). Hypothetical is "what if we tried X instead" (proposing an alternative). This skill is the former.
- If the topic has no clear subtractable component (rare — true of "transformer" but not of "softmax"), fall back to substitution or inversion.
Quick reference
- Input: topic + (optional) target-component-to-ablate.
- Output: one counterfactual framing statement + the concrete failure it reveals.
- Theory: Pearl's causal ladder — counterfactual reasoning illuminates causation in a way association cannot.
Related Skills
lyndonkl/conf-theme-clustering
testing
VerifiedTrustedCommunity
Cluster a conference's event records into a small set of coarse themes with finer sub-clusters, an explicit outlier bucket, and soft (multi-membership) affinities — using the hybrid embed-then-label pipeline (embed abstracts, reduce, density-cluster, then LLM-label the clusters) when embedding libraries are available, and an LLM-reasoned hierarchical fallback when they are not. Embeddings do the grouping; the LLM only names the groups. Conference-agnostic. Use when turning structured event records into a navigable theme map for preference elicitation and scheduling, when you need 6-8 reasonable themes rather than 20 muddy ones, or when overlapping talks must belong to more than one theme. Trigger keywords - theme clustering, cluster talks, embed then label, soft membership, outlier talks, conference themes, topic map.
127SKILL.mdUpdated Jun 28, 2026
lyndonkl/conf-schedule-optimization
development
VerifiedTrustedCommunity
Build a personal conference schedule as a constraint-optimization problem — hard constraints (no time overlap, room-to-room travel time, capacity/registration, the attendee's own must-attends and blackouts) plus a user-owned weighted objective trading interest against breadth, pacing (maximize contiguous free time), and serendipity. Surfaces unbreakable conflicts (two high-value overlapping talks the model cannot rank) as decisions for the human rather than silently picking, and reports what each choice traded away. Conference-agnostic. Use to turn a preference profile plus a theme map into a day-by-day plan, to resolve overlapping sessions, or to balance a packed vs paced schedule. Trigger keywords - schedule optimization, conference schedule, constraint optimization, overlapping talks, contiguous free time, conflict surfacing, packed vs paced.
127SKILL.mdUpdated Jun 28, 2026
lyndonkl/conf-program-extraction
development
VerifiedTrustedCommunity
Parse a heterogeneous conference program (markdown, HTML, PDF-derived text, or JSON) into normalized event records with per-field confidence scores and independent classification axes (topic, depth, format, prerequisites, recorded, capacity). Detects the program's format before extracting, treats every inferred field as uncertain (present vs inferred vs missing), and flags thin or missing abstracts so downstream enrichment can target them. Conference-agnostic. Use when ingesting a conference or event schedule into a structured store, normalizing a talk/session list, or extracting per-session metadata with calibrated confidence. Trigger keywords - program ingestion, parse schedule, session extraction, event records, conference program, talk metadata, per-field confidence.
127SKILL.mdUpdated Jun 28, 2026
lyndonkl/conf-preference-elicitation
development
VerifiedTrustedCommunity
Build a personalized preference profile from a small number of well-chosen, cluster-grounded questions instead of a long survey. Represents the person's interests as an uncertainty region over the theme map, picks the single highest-information-gain choice-based question (contrasting real talks from different clusters), balances exploiting known interests against exploring uncertain ones, deliberately injects outlier probes to fight selection bias, and stops as soon as the schedule would be stable. Also elicits the user-owned objective weights and hard constraints. Interactive — runs where it can actually ask the person. Conference-agnostic. Use to turn a theme map into a preference profile, to decide what to ask a conference attendee, or to elicit scheduling priorities. Trigger keywords - preference elicitation, ask few questions, information gain, choice-based questions, selection bias probe, objective weights, attendee preferences.
127SKILL.mdUpdated Jun 28, 2026