omarmohelal SecHelix for Claude Code
Read and follow skills/sechelix/SKILL.md as the canonical workflow. Use the project-local supporting resources under catalog/, references/, agents/, examples/, and scripts/.
Claude-specific orchestration guidance:
- Use subagents only for disjoint review lanes.
- Give each implementation/fix agent isolated ownership; do not let audit agents mutate the same release tree concurrently.
- Focused tests per lane; centralized heavy verification after integration.
- For High/Critical candidates, spawn an independent verifier that is asked to refute the finding rather than confirm it.
- Do not let dynamic testing leave authorized local/staging scope unless the operator explicitly authorizes a bounded production-safe check.
The methodology, severity rules, safety policy, and evidence standard are defined only in the canonical skills/sechelix/SKILL.md; do not fork them here.