skill-bibliothek für ai-agenten
Skills
Wiederverwendbare Fähigkeiten für Agenten. Suchen, filtern, kopieren.
Code review
Reviews a diff by severity, filtering low-confidence noise and verifying each bug is reachable before reporting it.
→ geprüfter code
Trustworthy tests
Writes tests that fail only when behavior breaks, mock only true external boundaries, and cannot silently pass on broken code.
→ vertrauenswürdig
Systematic debugging
Debugs by reproducing the failure, forming falsifiable hypotheses, and confirming root cause before changing any code.
→ grundursache behoben
Atomic commits
Splits a working diff into atomic, conventional commits with revert-safe, reviewer-ready messages.
→ saubere historie
Source verification
Runs multi-source research with adversarial fact-checking, tracing every claim to independent primary sources with calibrated confidence.
→ belegt
Trustworthy data analysis
Turns raw data into defensible conclusions by profiling and validating before analyzing, then stress-testing every number against statistical traps.
→ belastbar
Security review
Audits a diff for injection, authorization gaps, secret leaks, and unsafe deserialization by tracing untrusted input from source to sink.
→ abgesichert
Incident response
Drives live incidents mitigation-first through triage, roles, and status cadence, then reconstructs a blameless postmortem from evidence.
→ wiederhergestellt
Docs that get read
Chooses the right document type, then structures and edits it with Diátaxis and Google-style rules so docs stay readable and resist rot.
→ klarheit
Expensive-looking interfaces
Applies metric-based typography, spacing, hierarchy, and motion rules so an interface reads as intentional instead of generic.
→ wertig