| name | dillylang-invert |
| description | Applies Munger / Jacobi inversion on a problem statement. Trigger= /dillylang-invert PROBLEM |
Dillylang invert
[[THIS is_grounded_by: urn:unique_reference:dillylang::spec-primer]]
Apply Munger / Jacobi inversion on the given problem statement.
Stop asking "how succeed" and ask "what guarantees failure."
Prompt constraints
- Require concrete mechanism descriptions for every failure mode.
- Reject generic risks that lack a specific causal chain.
- Each failure mode must name the mechanism by which it causes damage.
Calibration example:
Rejected: "FM-1. Market changes — Likelihood: high, Severity: costly."
(No mechanism. How do market changes cause damage? Through what chain?)
Accepted: "FM-1. Price-sensitive users churn on first renewal —
Mechanism: free tier sets anchor price at zero; switching to paid
triggers loss aversion disproportionate to the dollar amount.
Likelihood: high. Severity: costly. Preventable by: usage-gated
tiers that establish value before the price conversation."
Output template
The problem statement from a Munger inversion point-of-view, e.g. "What would cause [desired state] to fail?"
Anti-goals (AG-n)
Desired-state inversions: what would I aim for if I wanted to fail?
Not causal chains (that's FM) — these are the goals of the adversary.
Failure modes (FM-n)
Causal chains: how does the damage happen, mechanistically?
Each entry must include:
- Mode: what goes wrong
- Mechanism: the specific causal chain — how it causes damage
- Likelihood: low | medium | high
- Severity: recoverable | costly | fatal
- Preventable by: what stops this
Near misses (NM-n)
Fragile-but-surviving conditions: things that almost fail but currently
hold. For each, name what currently prevents it from becoming a full
failure mode — the thing that would have to change for it to tip over.
n is sequential starting from 1 in each section.
Self-review
After generating your output, review each failure mode: state what
causal chain it reveals that the problem framing obscured. Failure
modes that name a generic risk without a specific mechanism have
not done work — strengthen or replace them.