Extensions
The first extension is sicario-guard.
Commands:
/sicario.init/sicario.assess/sicario.threatmodel/sicario.controls/sicario.evidence/sicario.verify/sicario.review/sicario.apply-findings
The extension is designed for Spec Kit command discovery. The deterministic
local implementation lives in the sicario CLI.
Hooks: executed vs. agent guidance
extension.yml registers hooks on after_specify, after_plan, and
after_tasks. These commands split into two honestly-labeled kinds:
| Hook command | Kind | What runs |
|---|---|---|
/sicario.verify | deterministic | sicario verify (the halting gate) |
/sicario.assess | deterministic | sicario assess |
/sicario.evidence | deterministic | evidence/assessment generation |
/sicario.threatmodel | agent guidance | a coding agent drafts the threat model |
/sicario.review | agent guidance | a coding agent reviews the change |
/sicario.controls | agent guidance | a coding agent updates control applicability |
/sicario.apply-findings | agent guidance | a coding agent turns findings into tasks |
sicario hooks reads .specify/extensions.yml, executes the deterministic
hooks, and reports the agent-guidance hooks (with a pointer to their command
doc) instead of pretending to run them:
sicario hooks # all events
sicario hooks --event after_tasks
The authoritative pass/fail decision is always the deterministic verify gate, not the AI-authoring steps.