Skip to content
Kitbay
Skill

security-and-hardening

MITUpdated on Nov 18, 2026118,985GitHub stars

A skill from Addy Osmani's agent-skills collection that treats security as a constraint on any change touching user data, authentication or outside systems.

What the skill teaches

A skill from Addy Osmani's agent-skills collection that treats security as a constraint on any change touching user data, authentication or outside systems. The agent starts with a threat model of trust boundaries, including model output, then follows three tiers of rules: things to always do, such as validating input, parameterising queries and hashing passwords; things to ask about first, such as new login flows or uploads; and things never to do, such as committing secrets or storing tokens in local storage. It covers sessions, headers, server-side request forgery, rate limits, dependency hygiene, privacy duties and risks specific to language model features, and ends with review and sign-off checklists.

Capabilities

  • Threat-model trust boundaries before coding
  • Apply always, ask-first and never rules
  • Harden sessions, headers and request handling
  • Review dependencies and privacy handling