Support claims with reasons
Otherdevflow/skills/cite
Asks for a brief explanation of each claim and a note where supporting evidence is missing.
Instructions and supporting material that help AI follow a consistent approach.
devflow/skills/cite
Asks for a brief explanation of each claim and a note where supporting evidence is missing.
devflow/skills/cot
Requests a clear explanation of the main solution steps and facts that can be checked.
devflow/skills/structured
Requests a JSON response from the model. Validate the output before processing it further.
devflow/skills/clarify
Asks the model to clarify important missing or ambiguous information before answering.
devflow/skills/safe
Provides guidance for briefly explaining requests that cannot be fulfilled and suggesting suitable alternatives.
lernraum/produktstrategie
A reasoned roadmap from feedback and metrics — the loudest request does not win.
kundenkompass/kundenservice-mit-eskalation
Answers grounded in the knowledge base — and escalation when the grounding is missing.
betriebsküche/prozessautomatisierung
Reliable flows instead of repetitive manual work — with idempotency and external verification.
codeschmiede/sicherheitsaudit
Finds and prioritises weaknesses in authorised systems — confirmed and potential kept apart.
produktlabor/saas-architektur
From idea to a secure, scalable, implementable architecture — including the failure cases, not just the happy path.
termekmuhely/feladatterv-ellenorzessel
Decomposes complex work, parallelises it, then verifies independently — it does not believe "done" just because a subtask says so.
devops-konyha/incidens-diagnosztika
Diagnoses production failures while minimising risk — and preserves evidence before intervening.
adatmuhely/adatelemzes-dontesig
Reliable analysis from raw data: quality first, assumptions recorded, correlation is not causation.
kutatolabor/melykutatas-ellenorzessel
Separates the proven, the estimated and the unknown — and never invents a citation.
kodkoznap/onallo-fejlesztes
Does not generate blindly: inspects, plans, writes the smallest coherent change, then actually runs the tests.
kanzlei-tools/widerspruch-melden
Gets the model to raise contradictions in your request instead of silently picking one reading.
promptmuhely/bizonytalansag-jelzes
Forces the model to separate knowledge from inference — and to say when it does not know.