Summary
Adds owasp_ast as a real schema field, replacing the current
crosswalk-file-only mapping to OWASP's Agentic Skills Top 10 (AST10)
with a proper, first-class field, the same treatment owasp_asi,
owasp_mcp, mitre_atlas, and nist_ai_rmf already get.
Current state
The AST10 mapping exists only in crosswalks/ave-to-ast10.json, a
separate file, not queryable the way the schema's other framework
fields are. README's own crosswalk table already lists this as planned:
"no dedicated schema field yet (owasp_ast planned)."
The crosswalk file's own coverage summary: 37 AVE records referenced
across 8 of AST10's 10 categories (AST08 and AST10 currently have no
mapped records at all in this file); 19 AVE records are explicitly
unmapped, mostly data-exfiltration/disclosure/standalone-injection
classes that don't fit AST10's scope, per the file's own
note_on_unmapped.
Scope
- Add
owasp_ast as an optional array field to
schema/ave-record-1.1.0.schema.json, pattern-matched to AST01
through AST10, same shape as the existing owasp_asi field.
- Additive, optional field: minor version bump only,
1.1.0 to 1.2.0,
per docs/specs/scaling-and-governance.md Section 2. No breaking
change to any existing record.
- Backfill
owasp_ast on the AVE records already listed in
crosswalks/ave-to-ast10.json's own mappings[].ave_ids arrays.
This file has no per-record confidence field (its confidence/rationale
lives in prose notes per AST category, a coarser grain than a
per-record flag), so the backfill is a direct read of its existing
mappings, not a confidence-filtered subset. Records the file lists
under gaps_in_ave or gaps_in_ast10 are explicitly not a mapping
and should not be backfilled.
- The crosswalk file itself stays, doesn't get deleted, it documents the
mapping rationale (including the gaps_in_ave/gaps_in_ast10 and
revision_note context) the schema field alone can't carry.
Out of scope
- Re-litigating which specific AVE records map to which AST10
categories, that's the crosswalk file's own job. This issue is about
adding the schema field and backfilling from the mappings as they
currently stand in crosswalks/ave-to-ast10.json, not redoing the
mapping work.
- A separate, independent AVE-to-AST10 crosswalk exists as an open PR
against the upstream OWASP AST10 project itself
(kenhuangus/agentic-skills-top-10#10), with its own confidence
grading (direct/partial) and an open CodeRabbit review comment on
one mapping (AVE-2026-00057 -> AST08) not yet addressed. That crosswalk
is a different file living in a different repository, not the one this
issue backfills from, and its outcome doesn't block or inform this
issue.
Summary
Adds
owasp_astas a real schema field, replacing the currentcrosswalk-file-only mapping to OWASP's Agentic Skills Top 10 (AST10)
with a proper, first-class field, the same treatment
owasp_asi,owasp_mcp,mitre_atlas, andnist_ai_rmfalready get.Current state
The AST10 mapping exists only in
crosswalks/ave-to-ast10.json, aseparate file, not queryable the way the schema's other framework
fields are. README's own crosswalk table already lists this as planned:
"no dedicated schema field yet (
owasp_astplanned)."The crosswalk file's own
coveragesummary: 37 AVE records referencedacross 8 of AST10's 10 categories (AST08 and AST10 currently have no
mapped records at all in this file); 19 AVE records are explicitly
unmapped, mostly data-exfiltration/disclosure/standalone-injection
classes that don't fit AST10's scope, per the file's own
note_on_unmapped.Scope
owasp_astas an optional array field toschema/ave-record-1.1.0.schema.json, pattern-matched toAST01through
AST10, same shape as the existingowasp_asifield.1.1.0to1.2.0,per
docs/specs/scaling-and-governance.mdSection 2. No breakingchange to any existing record.
owasp_aston the AVE records already listed incrosswalks/ave-to-ast10.json's ownmappings[].ave_idsarrays.This file has no per-record confidence field (its confidence/rationale
lives in prose
notesper AST category, a coarser grain than aper-record flag), so the backfill is a direct read of its existing
mappings, not a confidence-filtered subset. Records the file lists
under
gaps_in_aveorgaps_in_ast10are explicitly not a mappingand should not be backfilled.
mapping rationale (including the
gaps_in_ave/gaps_in_ast10andrevision_notecontext) the schema field alone can't carry.Out of scope
categories, that's the crosswalk file's own job. This issue is about
adding the schema field and backfilling from the mappings as they
currently stand in
crosswalks/ave-to-ast10.json, not redoing themapping work.
against the upstream OWASP AST10 project itself
(
kenhuangus/agentic-skills-top-10#10), with its own confidencegrading (
direct/partial) and an open CodeRabbit review comment onone mapping (AVE-2026-00057 -> AST08) not yet addressed. That crosswalk
is a different file living in a different repository, not the one this
issue backfills from, and its outcome doesn't block or inform this
issue.