Add cfgaudit → AVE crosswalk (static config-auditor)#67
Merged
Conversation
…veproject#64) Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: sonvisal <sonvisal10@gmail.com> Co-authored-by: Macbook <macbook@192.168.1.30>
…de (aveproject#66) Co-authored-by: sonvisal <sonvisal10@gmail.com>
cfgaudit (github.com/cfgaudit/cfgaudit) is a static auditor of committable AI-agent configuration files. It has 90 rules; 33 map onto 19 AVE behavioral classes (many-to-one, since cfgaudit slices threats by config surface where AVE slices by behavior). It emits each rule's primary AVE id in JSON and SARIF. The crosswalk also lists the config-file classes cfgaudit audits that carry no AVE class today (permission config, zero-click IDE auto-run, telemetry redirect, sandbox config, TLS-verification, container posture, MCP network posture, supply-chain pinning), so the taxonomy's coverage against a config-auditor is visible. Includes a cross-implementation validation: cfgaudit 1.9.0 and Bawbel Scanner 1.3.0, which share no code or ruleset and only the AVE taxonomy, run on the same SKILL.md files agree on the ave_id for 5 of the 10 shared-surface rules, with the divergences shown to be detection-pattern differences rather than mapping errors. Versions: cfgaudit 1.9.0, AVE 1.1.0, Bawbel Scanner 1.3.0.
Contributor
|
Tracking issue for the taxonomy gap list opened: #68 |
chaksaray
added a commit
that referenced
this pull request
Jul 23, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This adds a crosswalk from cfgaudit, a static auditor of committable AI-agent configuration files (Claude Code, Cursor, Copilot, Gemini, Codex, Devin, Zed, Continue), to AVE. It follows the
crosswalks/convention (.json+.md, alongside skillspector and clawscan).Versions: cfgaudit 1.9.0, AVE record set 1.1.0, Bawbel Scanner 1.3.0 (used in the validation below).
A different tool class from the two existing reference implementations. SkillSpector and clawscan are skill/MCP behavioral scanners. cfgaudit is a static config-surface auditor, so it exercises the taxonomy across a layer the other two do not touch. It emits each rule's primary
ave_idin JSON and SARIF (per yourave-in-sarif.md: the tool's own id stays the SARIFruleId,ave_idrides inproperties).Coverage. cfgaudit has 90 rules; 33 of them map onto 19 AVE behavioral classes (many-to-one, since cfgaudit slices threats by config surface where AVE slices by behavior). The other 57 rules check config-file classes AVE's skill and MCP-server records do not enumerate (permission/approval config, zero-click IDE auto-run, telemetry redirect, sandbox-config weakening, TLS-verification-disabled, container posture, MCP network posture, supply-chain pinning). The crosswalk lists these so the taxonomy's coverage against a config-auditor is visible.
Cross-implementation validation. cfgaudit and Bawbel Scanner 1.3.0, which share no code and no ruleset and only the AVE taxonomy, were run on the same
SKILL.mdfiles using cfgaudit's canonical trigger text unmodified (static engines only, no LLM). Of the 10 instruction-content rules that share a scan surface with Bawbel's file scan, both scanners independently emit the sameave_idon 5. The 5 divergences are detection-pattern differences, not mapping errors (Bawbel bundles a rule for each class; its pattern just did not match the same trigger). This is direct evidence the shared id interoperates across implementations.