Causal model construction
Skill ralfyishere/piensalo/micro-skills/causal-model-construction
PIÉNSALO is the open artificial cortex for AI—reduce context tokens, verify response quality, expand when needed, and safely fall back.
npx -y skills add ralfyishere/piensalo --skill causal-model-constructionAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 25 days oldThe repository was created 25 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Before intervening in a system, write the explicit causal chain from the proposed action to the intended outcome — every arrow marked verified or assumed — and check the weakest assumed arrow first. Activate when an intervention (fix, optimization, policy change, growth action) is proposed on a system whose mechanism has not been written down.
The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
2.6 KB, 447 tokens by cl100k_base, as published. Nobody here has run it
causal-model-construction
Trigger (observable): An intervention is proposed (fix, optimization, policy change, growth action) on a system whose mechanism has not been written down — the plan jumps from action to expected outcome with no stated path between.
When NOT to activate: The causal chain is already written and its arrows labeled this session; the action is trivially reversible and cheaper to try than to model; pure observation tasks with no intervention.
Procedure
- Write the chain explicitly: action → intermediate state(s) → intended outcome, as named nodes and arrows (X → Y → Z), not prose.
- Mark each arrow verified (observed/measured this session) or assumed (from memory, analogy, or hope).
- For each assumed arrow, name what else could produce the downstream node (confounders) and what else the action produces besides the intended effect (side paths).
- Identify the weakest assumed arrow — the one that, if false, makes the intervention useless or harmful — and check it before committing the intervention (this is where a discriminating test belongs).
- If the chain cannot be written, say so: an intervention with no articulable mechanism is a guess and gets labeled as one.
Required output
The written chain with per-arrow verified/assumed labels, named confounders/side paths, and the weakest arrow plus its check.
Verification
- The written chain exists and every arrow is labeled verified or assumed.
- The weakest assumed arrow is identified and checked (or its check is named and sequenced) before the intervention is endorsed.
Known risk: Modeling overhead on cheap reversible actions where trying is faster than theorizing. Mitigation: the not-when clause routes those to direct trial.
Max intended cost: ≤300 added output tokens; one check of the weakest arrow where feasible.
Evidence status: DESIGNED — specified from documented reasoning-failure modes; not yet executed as a packaged skill.
Lineage: Derived from a documented reasoning-failure mode — paraphrasing a system's behavior instead of stating its mechanism — with discriminating-test placement and instrument-distrust sequencing drawn from evidence-backed repair patterns.