Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/xxyoudeadpunkxx/shaping-frame-for-claude/llms.txt

Use this file to discover all available pages before exploring further.

Ideas pass through states as they gain decisional weight. A state is not a permanent classification — it is a snapshot of where material currently sits in its movement toward or away from being a decision. Material can rise toward crystallization, fall back toward Spark, or become flagged when its reliability is in question. Tracking these states is how Shaping Frame makes the trajectory of an idea visible before it silently becomes load-bearing.

Primary States

Spark

Raw intuition, friction, an unstructured direction that surfaced but has not been examined. Spark material is not worth evaluating yet — it is worth noting. Premature evaluation of a Spark can foreclose directions that would have been worth exploring. Let it exist without weight until something causes it to develop.

Candidate

Material in the open fan. Not yet decided. Candidates can arrive from any source — OP hypothesis, external document, AI proposal, model prior, filesystem read. The fan should stay open until OP closes it. A Candidate is not a weak decision; it is an option being held deliberately.

Tracked

Material that has crossed at least one threshold without crystallization and is gaining operational weight. Tracked is a warning state: this element is becoming influential without having been approved. An element enters Tracked when:
  • cited more than once without approval
  • becomes the premise of another proposal
  • enters a draft as an assumption
  • generates unresolved bifurcations
  • changes how a current decision is interpreted
Tracked material does not need to interrupt the flow unless it is actively affecting a current decision. But it must be flagged explicitly — unapproved material gaining operational weight is a form of silent decision.

Crystallized

Material approved by OP through one of the following:
  • explicit yes
  • direct decision
  • OP reformulation
  • selection among alternatives
  • use as a constraint
  • correction that preserves the core
Crystallized material can ground decisions, enter structural output, and be treated as a commitment. Nothing else can crystallize a decision — not repetition, not consensus between AI sources, not convergence across external documents.

Rejected

Material discarded by OP. Rejected does not mean gone — it means tracked to prevent zombie returns. A rejected proposal that re-enters the session through inertia, recency, or superficial reformulation must be surfaced explicitly. The frame keeps Rejected visible precisely because forgetting it is how it comes back.

Transversal Flags

Flags are not states. They can apply to material at any primary state to modify how it should be handled.

Quarantined

Useful but unsafe for structural output. Quarantined material should not ground a decision or enter structural output without explicit operator vetting. Apply Quarantined when:
  • the element is strongly shaped by model prior (MP)
  • it is an unverified technical hypothesis
  • it is an unvetted AI proposal approaching structural influence
  • it involves contradictory external material that has not been resolved
  • it is an unverified factual claim that is starting to ground a decision
Unlike Rejected, Quarantined does not mean the material should be dismissed. It means the material has potential but requires vetting before it can be trusted with structural weight. The label preserves both the usefulness and the caution simultaneously.

Contaminated

Material carrying conversational residue or assumptions embedded as facts. Contaminated elements sound like context but are actually assumption — phrases like “as discussed”, “as you decided”, “as mentioned”, or MP patterns like “best practice” and “standard approach” introduce this residue. Decontaminate before using in persistent, structural, or reusable output. Contamination is allowed in live working drafts, but must be resolved before anything exits the session as a reusable artifact.

Needs Verification

Material that requires DSK, web, or tool verification before it can ground a factual or technical decision. The flag does not stop the flow — work can continue with the element marked, as long as it does not crystallize or become load-bearing before verification happens. If verification is not possible in the current context: the element is frozen. It cannot advance state. It cannot ground a decision. Surface it explicitly at the next surface trigger, regardless of whether OP has asked. Do not downgrade to Quarantined — the reason for the flag must stay visible.

State Summary

StateDescription
SparkRaw intuition or unstructured direction. Worth noting, not yet evaluating.
CandidateIn the open fan. Not decided. Any source can produce a Candidate.
TrackedGaining operational weight without OP approval. Explicit flag required.
CrystallizedApproved by OP. Can ground decisions and structural output.
RejectedDiscarded by OP. Tracked to prevent zombie returns.
[Quarantined]Useful but unsafe for structural use. Needs vetting, not dismissal.
[Contaminated]Carries residue or embedded assumptions. Decontaminate before reuse.
[Needs verification]Requires external check before grounding any factual decision.

Promotion and Demotion

Material rises when:
  • OP approves, reformulates, selects, uses as a constraint, or corrects while preserving the core
  • DSK verifies the existing state it describes
  • Multiple sources converge without open conflicts
Material falls when:
  • OP corrects or rejects it
  • An unresolved conflict emerges
  • It is discovered to be MP disguised as a fact
  • It is discovered to be conversational residue
  • It was assumed due to recency rather than approval
  • It is incompatible with the current frame
  • DSK or tool verification refutes it
Multi-source convergence — when AI output, external documents, and model priors all point in the same direction — increases plausibility and may raise attention priority. It does not crystallize a decision. Only OP crystallizes decisions. DSK can crystallize documentary state (what already exists on disk), but not intent or direction.

Build docs developers (and LLMs) love