Decision snapshot builder
Skill hegu-1/claude-cognition-skills/skills/decision-snapshot-builder
Claude Code skills for the failure modes of long-horizon AI collaboration — continuity, judgment preservation, trap detection, faithful translation.
npx -y skills add hegu-1/claude-cognition-skills --skill decision-snapshot-builderAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 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
Build a clean decision surface from nonlinear or overloaded cognition. Use when the user has enough movement, material, or tension that a stable snapshot is needed before proceeding.
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
3.2 KB, as published. Nobody here has run it
Decision Snapshot Builder Skill
Version: 0.1
Purpose
This skill converts active cognitive movement into a stable decision snapshot.
A snapshot is not a full theory. It is a temporary structural crystallization that makes action, judgment, risk, and unresolved variables visible.
When to use
Use this skill when:
- the user has many ideas but no usable decision surface
- a conversation has become dense and needs crystallization
- execution is about to begin
- external communication requires a stable position
- the user needs to know what is decided and what is not
- continuation without a snapshot would increase distortion
Input expectations
Possible inputs:
- recent conversation segment
- current thread
- current phase
- active options or tensions
- known constraints
- trap signals
- intended next context such as meeting, email, execution, or reflection
Output contract
Return a structured snapshot with:
what_is_happeningwhat_matters_nowwhat_is_decidedwhat_is_not_decidedmain_constraintmain_riskrecommended_next_move
Example:
{
"what_is_happening": "the user is converging on a structural product framing",
"what_matters_now": "retain the original cognitive core while making it externally legible",
"what_is_decided": "the system should be framed as kernel-like infrastructure rather than a generic assistant",
"what_is_not_decided": "which deployment interface should be built first",
"main_constraint": "translation cannot erase the underlying structure",
"main_risk": "premature packaging into a standard product narrative",
"recommended_next_move": "write one outward-facing skill module before expanding the whole system"
}
Build logic
1. Detect the central tension
Find the actual decision surface, not just the topic.
2. Separate decided from undecided
Do not collapse uncertainty into fake resolution.
3. Surface constraint and risk
A useful snapshot must show what can break.
4. Keep the snapshot compact
The snapshot should reduce overload, not recreate it.
5. Preserve trajectory
A snapshot should not sever continuity. It should make the next move easier.
Snapshot style rules
A good snapshot should be:
- clear
- compact
- decision-relevant
- structurally faithful
- non-corporate in tone unless external translation is requested
Anti-patterns
Do not:
- turn the snapshot into a full PRD unless asked
- invent certainty
- mix decision, speculation, and implementation into one block
- oversmooth the user's actual judgment
- hide the unresolved part
Success criteria
This skill is working when:
- the user can suddenly see the shape of the situation
- next actions become easier
- ambiguity is reduced without flattening complexity
- the snapshot feels like a real temporary landing point