SKILL.md
Security and Safety Mindset
Intent
- Treat every change as a potential attack surface or failure amplifier.
- Ensure data classification, secret handling, and permission scopes stay compliant.
- Bake safety checks (rate limits, input validation, monitoring) into the design, not after.
Baseline Checklist
- Threat model quickly: Who could abuse this surface? What capabilities do they need? What happens if they succeed?
- Data stewardship: Classify data touched (PII, payments, assets) and enforce encryption, retention, and locality rules.
- Access + identity: Validate authn/authz paths, key rotation, wallet signatures, and privilege escalation barriers.
- Dependency hygiene: Pin versions, verify licenses, review changelogs, and prefer audited libraries/contracts.
- Secrets + config: Never log secrets; store them in the project’s approved secret manager. Guard env var usage.
Workflow
- Enumerate entry points (mobile UI, API, smart contract, admin tools) and list unchecked inputs.
- Define validation layers: schema-level, business-level, and environment-level (e.g., chain ID, platform version).
- Ensure every state change is reversible or compensatable (feature flags, contract pausing, migration guards).
- Instrument detection: structured logs, metrics, or on-chain events that can surface abuse or regressions fast.
- Document explicit “never do” actions (e.g., disable signature checks, bypass paywalls) inside the PR/issue notes.
Verification
- Run the project’s security/static analysis tooling (linters, contract analyzers, mobile scanners) and fix findings.
- Peer review the threat model summary; confirm secrets and keys are absent from diffs/logs.
- Validate abuse cases end-to-end (invalid payloads, replayed signatures, abusive traffic) before shipping.