Thunder Red Team
Adversarial security review of Thunderbolt's end-to-end encryption.
Install
npx skills add thunderbird/thunderbolt --skill thunder-red-team -g -a claude-code -y- Repository
- thunderbird/thunderbolt4.8k
- Category
- Security
- Updated
- 3d ago
- License
- MPL-2.0
- Context cost
- Mediumabout 3,000 tokens
- Runs scripts
- No, instructions only
What it does
Adversarial security review of Thunderbolt's end-to-end encryption. Attacks the security claims in docs/internals/architecture/e2ee-threat-model.md rather than reviewing code for correctness. Modes: a full sweep (`all`, or no argument) that fans every scoped pass out as a subagent then triages and refutes; a reproduce step that promotes a confirmed finding to an executed attack spec; a single read-only reasoning pass over one scope; or a live hunt against a booted local stack. Use when asked to "red team", "attack the encryption", "try to break E2EE", "run a red-team pass" or "run all red-team passes", or to validate a claim like "the server cannot read user data".
Red-teaming Thunderbolt E2EE
Ordinary review asks "is this code correct?" This asks "who breaks it, and how?" Your job is to falsify specific security claims. A run that ends with "the design looks sound" and no attempted attacks is a failed run.
Assume the design is wrong somewhere and go find where. Be adversarial, concrete, specific.
Confidence ladder
Every finding has a rung, the single most important thing to state about it. Never conflate them.
- L0 Hypothesis, an argument a vuln exists (a reasoning pass produces these).
- L1 Survives refutation, no guard found that kills it (a refuter subagent).
- L2 Reproduced by attack spec, executed proof against real client + backend, written as an
expected-failure (
test.fail()) so it stays green while the vuln is open and auto-trips when it is fixed. - L3 Live-reproduced, interactive, human-in-loop (the live-hunt track).
- L∞ Fixed & gated, the fix removes the
test.fail()tag; the spec becomes a permanent green regression gate that reds-out if the vuln ever returns.
Two rules that fall out of it: a passing spec (L2) is the confirmation, deterministic and durable,
not weaker than a live attack; and a failed live hunt does not refute a finding, it may only mean
the agent could not reach the adversary's position. Full rationale, the pipeline, and the pack contract
for other domains live in docs/internals/architecture/red-team-harness.md.
