feat: adversarial red team engine, custom persona builder, claim-level extraction, engine sweep#4
Merged
marceloceccon merged 3 commits intomainfrom Apr 25, 2026
Merged
Conversation
…l extraction, engine sweep - Adversarial Red Team engine: rotating attacker stress-tests positions before a parallel post-stress synthesis. Attacker confidence is out-of-band for the consensus formula; defenders run in parallel. - Custom persona builder: six-axis (risk, optimism, evidence, formality, verbosity, contrarian) slider UI. Server composes the system prompt from vetted phrase fragments — no user free-text reaches the LLM. - Claim-level disagreement extraction: post-final LLM pass emits structured contradictions with verbatim quotes per side. Quotes are verified against actual response content; fabrications are dropped. - Engine sweep mode: one-click sequential run across all three engines with a side-by-side comparison panel and explicit Cancel Sweep. Code-quality bundle: HMR-safe rate-limit cleanup, last-occurrence CONFIDENCE matching, snapshot usage reconstruction, judge/claim stream cleanup on cancel, hard-abort cost cap, tighter extractUsage type guards, fix for mockImplementation pollution across tests. Docs: README updated for the new engines and affordances; CHANGELOG.md and SECURITY.md added; newfeatures.md tracks per-feature rationale and QA notes. Tests: 207 -> 255 (+48 covering all new features and QA bundle).
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
…ntrolled format string' Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
before a parallel post-stress synthesis. Attacker confidence is
out-of-band for the consensus formula; defenders run in parallel.
verbosity, contrarian) slider UI. Server composes the system prompt
from vetted phrase fragments — no user free-text reaches the LLM.
structured contradictions with verbatim quotes per side. Quotes are
verified against actual response content; fabrications are dropped.
with a side-by-side comparison panel and explicit Cancel Sweep.
Code-quality bundle: HMR-safe rate-limit cleanup, last-occurrence
CONFIDENCE matching, snapshot usage reconstruction, judge/claim stream
cleanup on cancel, hard-abort cost cap, tighter extractUsage type
guards, fix for mockImplementation pollution across tests.
Docs: README updated for the new engines and affordances; CHANGELOG.md
and SECURITY.md added; newfeatures.md tracks per-feature rationale and
QA notes.
Tests: 207 -> 255 (+48 covering all new features and QA bundle).