Repository navigation
v1.0.0 — Initial release
Initial release
Adversarial review of functional specs before /speckit.plan locks in architecture — a Spec Kit extension complementary to /speckit.clarify (correctness) and /speckit.analyze (consistency).
What's in the box
extension.yml— extension metadata +speckit.red-team.runcommand registration + default config.commands/red-team.md— the command body (~24 KB): invocation parsing, preconditions, six-category trigger matching, propose-and-confirm lens selection, parallel adversary dispatch, findings aggregation, four-category resolution flow, hard-and-fast rule against editing historical SpecKit working records, full failure-mode handling.config-template.yml— lens catalog scaffold with two example lenses (Regulatory Adversary, Trust-Boundary Adversary) and inline schema documentation.README.md— install, adopt, invoke, troubleshoot.CHANGELOG.md,LICENSE(MIT),.gitignore.
Install
specify extension add --from https://github.com/ashbrener/spec-kit-red-teamOnce merged into the community catalog:
specify extension add red-teamDogfood-validated
Against a 500-line, 27-FR functional spec in a private project: 5 adversary agents dispatched in parallel returned 25 findings in ~1.5 min wall-clock. 19 of 25 findings met the meaningful-finding bar (severity ≥ HIGH AND represents an adversarial scenario /speckit.clarify / /speckit.analyze structurally cannot catch). One finding caught a cross-spec identifier-type drift between two halves of the same interface contract that had been introduced by a separate commit 1 hour earlier — a class of issue single-spec tools cannot surface.
Background
Originally proposed as a core command (github/spec-kit#2303); converted to an extension per maintainer guidance. Incorporates two rounds of Copilot review feedback (dot-notation alignment, CLI contract hygiene, error-message actionability, US spelling, threshold consistency).
Requires Spec Kit ≥ 0.7.0.