feat(ctf): Shadow Architect — admin portal tool description poisoning challenge#468
Open
steadhac wants to merge 1 commit intoGenAI-Security-Project:mainfrom
Open
Conversation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Adds
policy-bypass-shadow-architect— Expert/600pt challenge where the playerpoisons a tool description via the admin portal with a standing directive that
executes on every vendor conversation without further interaction.
Maps to the intentional CTF attack surface in
finbot/mcp/factory.py(_apply_tool_overrides) and
finbot/apps/admin/routes/api.py(update_tool_overrides).Files
finbot/ctf/definitions/challenges/policy_bypass/shadow_architect.yaml— challengedefinition, 3 tiered hints, prerequisite: rce-privilege-escalation, 600pts Expert
finbot/ctf/detectors/implementations/shadow_architect.py— ShadowArchitectDetector:fires when any tool on a monitored server has a poisoned description AND any tool
call succeeds on that server (handles directive in tool A causing tool B to fire)
finbot/static/js/admin/mcp-config.js— fix: read all textarea values at save timeto catch paste events that missed the input event listener
Labels
LLM01·LLM05·LLM06·CWE-94·CWE-284·CWE-693·AML.T0043·AML.T0051·ASI-01·ASI-02·ASI-03·ASI-06Test plan
reload_challenges.pyloads without errors