Optional AVE ids as SARIF rule tags, starting with SC8 and SC10? #930

Hi, I maintain AVE (Agentic Vulnerability Enumeration, https://github.com/aveproject/ave, Apache-2.0), an open catalog of behavioral vulnerability classes for agent components. I used an AI assistant to help draft this issue. Following CONTRIBUTING I am opening an issue first, and I will sign off any commit under the DCO if a PR is welcome.

I read SkillSpector v2.13.1 (the rules at https://github.com/NVIDIA/SkillSpector/blob/f3b25697d76256998ca9d42c198d256258d4c035/src/skillspector/nodes/analyzers/pattern_defaults.py, which pins the commit), rule by rule, and mapped its 78 rules in 18 categories to AVE records at class level. The result is a rule-level crosswalk pinned to that tag: https://github.com/aveproject/ave/blob/984a6f1378a1f1ad7c85329293caf4852de34934/crosswalks/skillspector-to-ave.json. Two examples from it:

1. SC8 (Shipped Python Bytecode) is close to AVE-2026-00075 (bytecode poisoning: a compiled .pyc references dangerous primitives that its sibling .py does not contain). I recorded it as "narrower", meaning the AVE class is a specific instance of what SC8 flags: SC8 fires on any shipped bytecode and does not test divergence from the source, so a hit is a lead and not a 00075 instance.
2. SC10 (Dependency Source Redirection) has no exact AVE counterpart that I can find. The nearest is AVE-2026-00073 (a static configuration value redirects a telemetry or API endpoint), which is about where an agent component sends traffic, not where a package manager resolves dependencies, so I left SC10 unmapped. That is a gap on the AVE side, and if you have a public reference for the behavior I would like to look at it.

The question: would you accept optional AVE ids as SARIF rule tags for the patterns with a class-level match, as a small data file maintained on our side with a CI check that referenced ids are still active? Docs-only is fine too, and so is no. Mappings are class-level equivalence, not identity.

No endorsement is implied either way, and only one other scanner (cfgaudit) emits AVE ids today. The crosswalk is pinned to a release tag and will go stale as SkillSpector changes, which is why I would rather it lived next to your rules than in our repo.