Repository navigation
feat(mcp): expose the security lens as an audit_attack_surface tool - #75
Merged
Merged
Conversation
Closes #64. The MCP server offered only extension-point detection; it now also maps where untrusted code or data can enter and execute, so an AI client can ask both questions about a local checkout without leaving the conversation. The tool returns exactly the payload `modscan-audit --json` emits, built by calling that same renderer rather than reassembling it. That is deliberate: it means the two cannot drift, and in particular that the non-coverage disclaimer travels with the data. A client reading the result cannot mistake it for "no vulnerabilities found", and an empty result still says so explicitly - covered by its own test. The two tools keep separate vocabulary, as the lenses do everywhere else: moddability answers carry a score and signals, security answers carry severity, confidence and an MS-SEC id, and neither is phrased in the other's terms. A test asserts no moddability score leaks into a security payload. Offline and free, like the detection tool: no LLM, no code execution. Suite 25/25; benchmark unchanged at 9/20. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
MODScan attack-surface diff —
|
MODScan extension-point diffNo breaking changes to extension points Added (new extension points)
|
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
Closes #64. The MCP server offered only extension-point detection; it now also maps where untrusted code can enter and execute, so an AI client can ask both questions about a local checkout.
Built through the CLI's own renderer
audit_attack_surfacereturns exactly the payloadmodscan-audit --jsonemits — by calling that renderer, not by reassembling it. Deliberate: the two cannot drift, and specifically the non-coverage disclaimer travels with the data.A client cannot read the result as "no vulnerabilities found", and an empty result still carries the disclaimer — covered by its own test, because that's the case where a false sense of security is most likely.
Separation held
The two tools keep their own vocabulary, as the lenses do everywhere else:
score,signalsseverity,confidence,MS-SEC-*idA test asserts no moddability score leaks into a security payload.
Test plan
tests/test_mcp_server.py: payload shape,MS-SEC-PICKLE+MS-SEC-SUBPROCESSfound, ranked most-dangerous-first, literalshell=Trueelevated to high,limithonoured, no moddability vocabulary in the payload, and the empty-result disclaimer.mcpextra (unchanged pattern).🤖 Generated with Claude Code