open/llm-security-commons
1
0
Fork 0
Code Issues Releases 1 Activity
llm-security-commons/conformance
History
Exact
Kjell Tore Guttormsen 8f08e9ac73 feat(carriers): three cases minted, and the id is named rather than adopted 
Both runtimes answered the 2026-08-11 decision request, and they did not ask
for the same name. The guard ruled `sanitize:` names its `Finding.detector`
and offered all three labels for verbatim adoption. llm-security, asked
directly, answered that the name would make its conformance result read as a
claim about neutralisation it does not perform.

Two things decided it. The guard's own unprompted correction: prefix ==
detector holds for those six labels and is no general law in its runtime
(`egress:*` carries detector="output"; decode-and-rescan yields two-part
`decoded:lexicon:*`). A prefix whose meaning is recoverable only by reading
one implementation cannot carry a shared id space. And a measurement taken
here at be9759b: on the surface the guard's own ruling pinned, `sanitize()`
returns changed text on all three carriers, so the counterargument's decisive
case -- that `scan_output` mutates nothing -- does not reach this surface.

Not a mediation. Neither runtime claimed the shared id must equal its label,
and `override:ignore-previous` already carries two different alias strings.

- carriers.json 0.1.0 -> 0.2.0: carrier:zero-width / :bidi-override /
  :unicode-tag, aliased to the guard's labels. No aliases.llm_security --
  that runtime's carrier findings carry no id yet, and publishing the alias
  is the irreversible act that forces the table into its declared set.
- manifest 0.3.4 -> 0.4.0: entry_points_by_scope, carrying findings accessor
  and fixture presentation per scope per runtime. This was objection (c), and
  it blocked minting harder than the name did.
- Corpus 90 -> 93. Measured through sanitize(text, source=Source.INPUT) at
  guard v0.7.0; verified by a separate checker that re-derives everything from
  disk -- a generator agreeing with itself proves nothing.
- CLAUDE.md gains the two rules that are not derivable from the data: a shared
  id space cannot rest on a one-runtime prefix, and publishing an alias -- not
  minting the case -- is the irreversible act.

Not minted on purpose: no artifact-side id (the other runtime would only fail
them), and no ZWJ-exemption case (U+200D between two emoji is exempt on both
guard surfaces since v0.6.1; the fixture avoids it rather than trips it).

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01U3o4zSQ2kwLsgJNU7apK3Z
2026-08-13 21:17:44 +02:00
..
active__autolink feat(conformance): ship the six active-content cases; the id space already existed 2026-08-10 21:15:25 +02:00
active__data-uri feat(conformance): ship the six active-content cases; the id space already existed 2026-08-10 21:15:25 +02:00
active__markdown-image feat(conformance): ship the six active-content cases; the id space already existed 2026-08-10 21:15:25 +02:00
active__markdown-link feat(conformance): ship the six active-content cases; the id space already existed 2026-08-10 21:15:25 +02:00
active__raw-html feat(conformance): ship the six active-content cases; the id space already existed 2026-08-10 21:15:25 +02:00
active__reference-link feat(conformance): ship the six active-content cases; the id space already existed 2026-08-10 21:15:25 +02:00
carrier__bidi-override feat(carriers): three cases minted, and the id is named rather than adopted 2026-08-13 21:17:44 +02:00
carrier__unicode-tag feat(carriers): three cases minted, and the id is named rather than adopted 2026-08-13 21:17:44 +02:00
carrier__zero-width feat(carriers): three cases minted, and the id is named rather than adopted 2026-08-13 21:17:44 +02:00
config__bypass-safety fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
config__disable-filtering fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
config__ignore-claudemd fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
config__override-system-prompt fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
data-uri__executable fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
evasion__academic-context fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
evasion__educational fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
evasion__fictional fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
evasion__hypothetical fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
evasion__red-team fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
evasion__research-purpose fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
evasion__security-audit-sim fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
evasion__test-pretense fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
evasion__thought-experiment fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hidden-comment__agent-directed fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hitl-trap__approve-immediately fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hitl-trap__brevity-skip fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hitl-trap__details-omitted fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hitl-trap__just-approve fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hitl-trap__no-need-review fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hitl-trap__scope-minimization fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hitl-trap__suppress-review fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hitl-trap__too-long fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
homoglyph__cyrillic-latin-mix fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
html-obfuscation__aria-label fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
html-obfuscation__display-none fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
html-obfuscation__offscreen fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
html-obfuscation__visibility-hidden fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
html-obfuscation__zero-font fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
html-obfuscation__zero-opacity fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
html-obfuscation__zero-size-overflow fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hybrid-p2sql__inject-sql fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hybrid-p2sql__sql-inject fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hybrid-recursive__conditional-response fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hybrid-recursive__inject-output fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hybrid-xss__event-handler fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hybrid-xss__iframe-src fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hybrid-xss__javascript-uri fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hybrid-xss__script-tag fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
hybrid-xss__script-tag--src-no-close feat(conformance): the witness case, and the derivation rule that had no room for it 2026-08-11 13:39:19 +02:00
identity__new-directive fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
identity__new-role fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
identity__no-longer fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
identity__pretend fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
identity__you-are-now fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
indirect__direct-address fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
indirect__instruction-to-ai fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
indirect__trigger-based fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
leetspeak__disregard fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
leetspeak__forget-instructions fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
leetspeak__ignore-previous fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
leetspeak__override fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
markdown__link-anchor-injection fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
markdown__link-ref-comment fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
multi-lang__french fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
multi-lang__german fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
multi-lang__spanish fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
nl-indirection__download-execute fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
nl-indirection__extract-exfiltrate fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
nl-indirection__fetch-execute fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
nl-indirection__read-dotfile fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
nl-indirection__send-external fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
normalization__claudemd-conflict fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
normalization__prioritize-user fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
normalization__skip-confirmation fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
normalization__user-precedence fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
override__disregard fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
override__forget-instructions fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
override__ignore-previous fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
override__override-instructions fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
override__prior-constraints fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
override__suspend-constraints fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
reframing__actual-task fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
reframing__real-purpose fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
reframing__should-actually-do fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
spoofed-header__hash-system fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
spoofed-header__inst fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
spoofed-header__instructions fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
spoofed-header__pipe-system fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
spoofed-tag__system-close fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
spoofed-tag__system-open fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
sub-agent__delegate-bypass fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
sub-agent__spawn-dangerous fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
unicode__zero-width-in-word fix(conformance): name the fixture field pattern_id, not id — the schema already owns that word 2026-08-10 04:44:59 +02:00
manifest.json feat(carriers): three cases minted, and the id is named rather than adopted 2026-08-13 21:17:44 +02:00