Skip to content

Align ASI codes and titles with the published Top 10 for Agentic Applications 2026 v1.0 - #11

Merged
hackwither merged 1 commit into
hackwither:mainfrom
avp9-nexus:asi-v1-alignment
Sep 11, 2026
Merged

hackwither merged 1 commit into
hackwither:mainfrom
avp9-nexus:asi-v1-alignment

Conversation

@avp9-nexus

Copy link
Copy Markdown
Contributor

Closes #10.

ASITitles in internal/report/report.go followed an earlier draft of the numbering (ASI04 to ASI09 carried different titles), and every asi_refs was coded against that draft. This updates the table to the published v1.0 list and renumbers the affected checks so each keeps the meaning it was given.

Table

  • ASITitles: the ten published v1.0 titles.
  • New TestASITitles_MatchPublishedV1 pins them. It fails against the previous table; a future change to the standard has to change the test on purpose.

Codes, intent preserved

  • http-rate-limit-absence: ASI06 to ASI08 (Cascading Failures). TestRateLimitProbe updated accordingly.
  • transport-plaintext, transport-downgrade: ASI04 to ASI07 (Insecure Inter-Agent Communication, which is what the draft's ASI04 meant).
  • mcp-tmpl-high-risk-tool-names: ASI07 to ASI05 (Unexpected Code Execution); ASI02 kept.

Docs

  • docs/ASI_MAPPING.md: rows updated, the template added to the table, and the two "corrections" paragraphs now name the draft and v1.0 codes side by side, with a third paragraph recording the table drift itself.

Left alone on purpose

The four checks citing ASI09 (tls-cert-health, mcp-tool-capability-surface, mcp-dynamic-dispatch, mcp-instructions-exposure) recorded observability facts under a draft category the final list dropped. Whether to uncite them like mcp-auth-posture or keep ASI09 is your call; the mapping doc says so rather than deciding.

Checks

gofmt -l empty, go vet ./..., go build ./..., go test -race ./... all pass on go1.22.12. The two new or changed assertions were confirmed to fail against the previous report.go and transport/checks.go respectively before the fix was applied, so they are not passing by accident.

…ications 2026 v1.0

ASITitles in internal/report/report.go followed an earlier draft of the
numbering (ASI04 to ASI09 carried different titles), and every asi_refs
was coded against that draft. Update the table to the published v1.0
list, pin it with TestASITitles_MatchPublishedV1, and renumber the
affected checks so each keeps the meaning it was given:

- http-rate-limit-absence: ASI06 -> ASI08 (Cascading Failures);
  TestRateLimitProbe updated.
- transport-plaintext, transport-downgrade: ASI04 -> ASI07 (Insecure
  Inter-Agent Communication, which the draft's ASI04 meant).
- mcp-tmpl-high-risk-tool-names: ASI07 -> ASI05 (Unexpected Code
  Execution); ASI02 kept.

docs/ASI_MAPPING.md: rows updated, template added, the two corrections
now name draft and v1.0 codes side by side, plus a note on the table
drift. The four ASI09 checks are left as they are on purpose.

Refs hackwither#10.
@hackwither

Copy link
Copy Markdown
Owner

hey looks good to me. thanks for the detailed description on your contribution

@hackwither
hackwither merged commit 3eef6ef into hackwither:main Sep 11, 2026
3 checks passed
@avp9-nexus

Copy link
Copy Markdown
Contributor Author

Thanks for the review and the merge, glad the description was useful.

One point the PR left to you is still open on main: tls-cert-health, mcp-tool-capability-surface, mcp-dynamic-dispatch and mcp-instructions-exposure still cite ASI09, which is Human-Agent Trust Exploitation in v1.0; they were coded against the draft's ASI09, Observability & Auditability Gaps, a category the final list dropped. Tell me which way you prefer, uncited like mcp-auth-posture or mapped to a v1.0 code, and I'll send a small follow-up PR.

Avp9.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

ASI category codes follow an earlier draft, not the published Top 10 for Agentic Applications 2026 v1.0

2 participants