feat(audit): INJ1 prompt-injection signature detector#49
Merged
Conversation
…HARDENING §2.2.6) Deterministic scan of tracked text for invisible / bidirectional control characters — the near-zero-false-positive signal of injection / homoglyph smuggling. Mirrors the boundary-detector shape (detect_inj1 + build_injection_detectors), wired into the runner as deprecated=True so it is skipped by the default gate (opt-in only) and cannot red the fleet-wide audit. Reports codepoint+position only (D-INJ-3); content-marker opt-out; \u escapes so it never self-triggers. 7 INJ1 tests; full suite 1126 passed. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
The last Phase-1 item of OperationsCenter's Harness Trust-Hardening (§2.2.6) — the outer layer of the injection defense (the load-bearing control is OC's code-computed typed verdict, already merged in OC #349/#350).
detect_inj1deterministically scans tracked text for invisible / bidirectional control characters — the unambiguous, near-zero-false-positive signal of injection / homoglyph smuggling (hidden instructions aimed at a model or human reviewer).Safety (designed not to break the fleet-wide audit)
detect_inj1(context) -> DetectorResult+build_injection_detectors()), wired into the runner.deprecated=True→ skipped by the default gate (opt-in via--only INJ1 --include-deprecated). A repo's own injection-handling code can't red the audit.custodian:allow-invisible-charscontent marker.\uescapes throughout → the detector never self-triggers.Verification
7 INJ1 tests; full Custodian suite 1126 passed; self-audit clean; ruff clean.