Public evidence ledger

Public evidence and limitations ledger · Operations Web Function 0.1.0.

Download the complete plugin · Source SHA-256: 71565a113519ccf04bb1d47ecf537f7a49701d9001913551fa92528986709236

This ledger supports narrow claims with dated evidence. It does not claim to be the “most authoritative.” Authority is earned through reproducible breadth, freshness, corrections, independent use, and citations.

Claim status

VERIFIED — the cited evidence directly supports the claim within its stated scope.
PROVISIONAL — evidence exists but requires another vantage, a longer window, or public reproduction.
UNKNOWN — the required observation has not been completed.
RETRACTED — the claim was wrong; the correction and reason remain visible.
SUPERSEDED — a later claim or method replaces it.

Evidence entries

fleet-denominator-2026-08-13

Claim: One bounded external fleet run checked all 221 selected domains and recorded an honest denominator and aggregate classifications.
Status: VERIFIED
Scope: 220 private-roster domains plus one external control; one observation window.
Evidence tier: E2 — multiple identities/canary with source and roster hashes in the private immutable receipt.
Result: 197 OK, 0 infected, 2 suspect, 2 down, 0 blocked, 20 unreachable.
Public artifact: examples/receipts/fleet-aggregate-2026-08-13.json
Redaction: property identities, domain-level findings, paths, and raw responses withheld.
Limitation: this was a bounded public guard, not the full field-hardened internal monitor, and it did not inspect authenticated layers.
Next action: publish a stable probe interface and make the current tested internal engine implement it.

detector-false-positive-correction-2026-07-21

Claim: A security detector's false positives were converted into boundary-aware parsing, positive and negative regression fixtures, and a larger test suite.
Status: VERIFIED
Scope: one monitor family and its dated correction history.
Evidence tier: E3 — source, append-only run log, and 116 local assertions reviewed in the authoring workspace.
Public artifact: examples/case-studies/02-monitor-false-positive.md
Limitation: private site identities, authenticated configuration, and baseline contents are not distributed.
Next action: publish the sanitized detector fixture and generated test receipt in a later release.

unknown-is-not-clean-2026-08-10

Claim: A corrected access audit separated unresearched/blocked observations from verified failures instead of scoring missing evidence as clean.
Status: VERIFIED
Scope: 285 properties in a dated internal audit; 21 remained explicit unknowns.
Evidence tier: E3 — authenticated aggregate audit plus regression tests.
Public artifact: examples/case-studies/03-unknown-not-clean.md
Limitation: property rows, credential coverage, account identifiers, and client names remain private.
Next action: publish a fully synthetic audit fixture that proves the state partition and residual assertion.

skills-need-citable-pages-2026-08-02

Claim: Twenty-four skill files and ZIP entries were not independently citable until the publishing system generated one provenance-bearing implementation page per skill and extended the runner and verifier to keep the page tier current.
Status: VERIFIED
Scope: the skill-pack publication system at the dated release.
Evidence tier: E3 — generator, runner contract, verifier, and learning record reviewed.
Public artifact: examples/case-studies/04-skills-need-pages.md
Limitation: later page availability must be reverified separately; historical generation is not proof of current reachability.
Next action: give every Web Function skill a clearly classified implementation page and verify task ↔ hub ↔ skill ↔ page ↔ release consistency.

Open evidence gaps

Independent third-party reproductions of the downloadable Web Function pack: UNKNOWN.
Independent citations to the Web Function methodology: UNKNOWN until observed and linked.
Package installed/enabled in the canonical marketplace: UNKNOWN; this output is a delivery artifact, not an installation receipt.
Candidate skills merged into the canonical marketplace and three candidate bounded tasks registered in the Task Library: UNKNOWN; registry/capability-map.json is an intake plan, not a merge or registration receipt.
Every existing production monitor reconciled to one canonical decision engine: UNKNOWN; current copies and schedules conflict.
One authoritative property registry across hosting, publishing, measurement, and access projections: UNKNOWN; current denominators differ by purpose and freshness.
Current public page and download availability from an independent browser after publication: pending read-back.

Required fields for future entries

Every new entry records: stable evidence ID; narrow claim; status; subjects and denominator; observation window; observed and reverified dates; evidence tier; collector and version/hash; positive and negative controls; metrics; limitations; public, sanitized, and private artifact references; independent reproduction; redaction class; related incident, run, skills, and job; owner; next action; and supersession link.

This page is generated from docs/EVIDENCE-LEDGER.md. Do not hand-edit the public copy.