ResultRST-2026-0202v0.1

Pooled F1–F4: high warnings on every surface — README 3.4%, agent instruction 3.0%, governance 0.9%

Primary rule, 236 repositories: README 8/233, agent instruction 7/236, governance 2/235. Rounds swung from one high-warning cell to nine. The company panel's concentration on agent instruction files did not recur; the concrete-path kind (L3) still leans toward them.

Research status: STABLE the current conclusions are relatively stable
Evidence level: E2 Controlled experiment
Result: MIXED
Data basis: BLIND PUBLIC DATA Real public files measured by an algorithm inside a sealed process. No person and no AI model read them; only coded features left the process. The numbers may be distorted.
Version: 0.1
Updated: 2026-10-13
Created: 2026-10-13
Domain: Privacy
Program: PRG-2026-0201 AI Agent Publication Boundary and Security Closure (APBSC)
Authors: Neo.K (EveMissLab)
AI collaborators: Splice (Claude Code, Anthropic) — Phase II protocols, runners, audits, reports and the integrated edition

Observed result

metrics

by_surface

GENERAL_README

high_warning: 8
analyzable: 233
share: 0.0343
ci95: 0.0149
0.0665
l3: 2

AI_INSTRUCTION_SELECTED

high_warning: 7
analyzable: 236
share: 0.0297
ci95: 0.012
0.0602
l3: 4

GOVERNANCE_FAMILY

high_warning: 2
analyzable: 235
share: 0.0085
ci95: 0.001
0.0304
l3: 1

paired

repositories

236

fully_analyzable

232

high_warning_pattern

AI_INSTRUCTION_SELECTED: 5
GENERAL_README: 7
GENERAL_README+AI_INSTRUCTION_SELECTED: 1
GOVERNANCE_FAMILY: 2
NONE: 217

instruction_vs_readme

only_AI_INSTRUCTION_SELECTED: 5
only_GENERAL_README: 7

instruction_vs_governance

only_AI_INSTRUCTION_SELECTED: 6
only_GOVERNANCE_FAMILY: 2

repositories_by_owner_type

Organization: 70
User: 166

Interpretation

interpretation: High-warning context exists on every surface; surfaces are not separated by the intervals; any single round is unrepresentative.

Limitations

limitations: Exploratory, non-probability frame; not a prevalence estimate.
No human checked any finding; secret-like means a shape and was never tested.
Current public state only, observed once.

Claims

Source	Relation	Target	Status	ID
`RST-2026-0202` Pooled F1–F4: high warnings on every surface — README 3.4%, agent instruction 3.0%, governance 0.9%	`supports`	`OBS-2026-0208` O8 — Agent-related publication-boundary signals appear among individuals, organizations and companies alike	`ACTIVE`	`REL-2026-0594`
`RST-2026-0202` Pooled F1–F4: high warnings on every surface — README 3.4%, agent instruction 3.0%, governance 0.9%	`qualifies`	`OBS-2026-0205` O5 — In a frozen 60-repository company panel, both high-warning results were on agent instruction files	`ACTIVE`	`REL-2026-0595`

Relations

Source	Relation	Target	Status	ID
`RST-2026-0202` Pooled F1–F4: high warnings on every surface — README 3.4%, agent instruction 3.0%, governance 0.9%	`supports`	`OBS-2026-0208` O8 — Agent-related publication-boundary signals appear among individuals, organizations and companies alike	`ACTIVE`	`REL-2026-0594`
`RST-2026-0202` Pooled F1–F4: high warnings on every surface — README 3.4%, agent instruction 3.0%, governance 0.9%	`qualifies`	`OBS-2026-0205` O5 — In a frozen 60-repository company panel, both high-warning results were on agent instruction files	`ACTIVE`	`REL-2026-0595`
`EXP-2026-0202` Rounds F1–F4 — fresh surface-balanced samples (236 repositories, 2026-10-13)	`produces`	`RST-2026-0202` Pooled F1–F4: high warnings on every surface — README 3.4%, agent instruction 3.0%, governance 0.9%	`ACTIVE`	`REL-2026-0593`

History and provenance

Canonical URL

https://evemisslab.com/ai/results/RST-2026-0202/

Machine-readable

/ai/results/RST-2026-0202/index.json

Snapshot

AI-SNAPSHOT-v0.1-58e716495670

Provenance

source: EveMissLab research collection: AI Agent Publication Boundary and Security Closure (真本體論14)
extracted_by: Splice (Claude Code), reading the open-source release's own manifests, locks and pooled outputs, and the hashes of the internal drafts
extracted_at: 2026-10-13
generator: tools/extract_all.py
claim_boundary: status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports