Collaboration pattern
Skill zhqi3/collaboration-pattern/skill/collaboration-pattern
Use when work reaches a material uncertainty or decision boundary, including unclear goals, unverified facts, customer assumptions, contradictory evidence, changing scope, completion claims, or costly hard-to-reverse choices.From its SKILL.md
npx -y skills add zhqi3/collaboration-pattern --skill collaboration-patternAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 26 days oldThe repository was created 26 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.
- 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 file declares
Copied from the file, not written here
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
5.8 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it
Collaboration Pattern
Overview
Use this pattern when missing or contradictory information could change an important decision. It helps an AI collaborator distinguish what must be checked now from what can be handled with a small reversible choice.
Do not turn every ambiguity into an investigation. The question is whether the unknown can materially change scope, architecture, data shape, validation, cost, safety, or reversibility.
Decision Loop
Understand the current decision
-> identify unknowns that could change it
-> decide whether a factual probe is required
-> choose an appropriate specialist capability
-> act under that capability's rules
-> verify material claims
-> reconsider when evidence changes
Map Only Decision-Relevant Uncertainty
- Known knowns: confirmed goals, evidence, constraints, and decisions.
- Known unknowns: open questions already recognized.
- Unknown knowns: tacit preferences, standards, workflow knowledge, or customer language that examples, observation, prototypes, or interviews can reveal.
- Unknown unknowns: hidden constraints, unenumerated behavior, alternatives, edge cases, and blind spots.
This map is a thinking aid, not a mandatory report. Show only distinctions that change the next step.
Apply the Probe Gate
REQUIRED REFERENCE: Read references/probe-protocol.md when a probe triggers.
Probe before relying on a claim when:
- a decision depends on an unverified file set, schema, API contract, source set, runtime behavior, or customer workflow;
- an error could materially change scope, architecture, data shape, validation, cost, safety, or reversibility;
- an acceptance criterion needs exact values that have not been observed;
- new evidence contradicts a load-bearing premise; or
- a material blind spot can be checked cheaply against reality.
Use a labeled conservative default when the unknown is small, reversible, and isolated. Do not probe when checking would cost more than making and clearly labeling that choice.
A load-bearing claim without enough evidence remains UNKNOWN. Do not choose, recommend, provisionally select, or claim completion when that choice depends on the unresolved claim. Define only a reversible information-gathering step that does not commit the decision.
Route by Capability
| Situation | Portable capability |
|---|---|
| Tacit or unclear needs | Discovery or design exploration |
| Unknown customer workflow or demand | Interviews, observation, and purchasing evidence |
| Agreed direction | Implementation planning |
| Unexpected failure | Systematic diagnosis |
| Substantial work appears complete | Independent review and verification |
| Claims depend on external facts | Research and source verification |
Read references/specialist-routing.md for small fallbacks and optional client mappings.
If the needed specialist workflow is available, use it under its own rules. If it is absent, name the missing capability and use the documented minimum fallback. Stop when the missing capability makes a defensible decision impossible. Never claim that an unavailable workflow ran.
Do not reproduce complete planning, debugging, research, interviewing, or review systems inside this Skill.
Reconsider Existing Decisions
Approval does not make a contradicted premise true. When evidence changes:
- name the premise that changed;
- identify the requirements, tests, plans, or decisions that depend on it;
- preserve unaffected work;
- return affected work to probing or design;
- verify the revised decision before continuing.
When customer workflow, buyer, budget, alternatives, or willingness-to-pay evidence is load-bearing and absent, keep the product definition unresolved. A neutral problem hypothesis and discovery plan are allowed; a settled feature set or roadmap is not.
For all, none, complete, or similar set-wide claims, enumerate the claimed set and check every relevant member. If a wider set is unknown, report exactly the enumerated or observed members and leave the wider status unresolved. A bounded completion claim is allowed only for an explicitly identified subset whose every member was verified. Do not claim the wider set complete or invent after-the-fact labels such as agreed or in-scope to manufacture a subset.
Output Contract
Default to a short decision-focused result:
Current decision:
Material unknowns:
Probe required: yes/no — why
Clarify now:
Handle conservatively:
Recommended capability:
Omit fields that do not help the next decision. Use the full uncertainty map only when it improves the choice.
Common Mistakes
| Mistake | Correction |
|---|---|
| Investigating every small ambiguity | Choose and label a reversible default |
| Designing from a plausible schema | Inspect a real instance first |
| Turning an idea into an MVP without customer evidence | Keep a problem hypothesis and gather workflow/purchasing evidence |
| Preserving a plan after its premise fails | Reopen only the affected decisions |
| Naming a specialist that is not installed | Use a documented fallback or stop |
| Claiming completion from samples | Enumerate the claimed set; bound completion to fully verified members |
References
references/probe-protocol.md— enumerate-and-verify procedure.references/specialist-routing.md— portable fallbacks and optional client mappings.references/example-prompts.md— concise prompts for common decision boundaries.
What ships with it: 4 files
8.2 KB alongside SKILL.md
agents/
- openai.yaml273 B
references/
- example-prompts.md1.9 KB
- probe-protocol.md3.8 KB
- specialist-routing.md2.2 KB