Risk Review · Payload v1

Use each vector to guide manual verification of the underlying findings.

Payload schema: 1.0.0 · Engine: 2.0.0 · Scoring: heuristic-density-v1 · Scan date: 2026-10-31.

Cognitive Ownership Risk 53.7/100 · HIGH

Can maintainers explain intent, assumptions, and trade-offs? 16 finding(s).

Hallucinated Dependency & Supply Chain Risk 0.0/100 · LOW

Are imports declared and package/API names verified? 0 finding(s).

Agent Over-Engineering Ratio 0.0/100 · LOW

Are wrappers, guards, and repeated structures necessary? 0 finding(s).

PR Review Velocity Impact 60.3/100 · HIGH

Which changes need focused human review? 16 finding(s).

What These Metrics Mean

Adjusted Signal: Legacy pattern score normalized per 1,000 scanned lines and multiplied by the repository-level time factor. It is not an estimate of AI authorship.
Raw Pattern Score: Sum of severity- and context-weighted pattern findings before applying the time factor.
Time Factor: A coarse multiplier based on repository last-push metadata. It does not identify when individual code was written.
Pattern Hits: Count of configured pattern findings. Review each location and surrounding code before acting.
HC Hit Rate: High- and critical-severity findings per scanned file. Severity is a configured detector weight, not proof of a vulnerability or authorship.
Lines / Files: Scan coverage denominators used by the legacy normalized score.

Pattern Findings

The scanner identified 16 distinct pattern matches across 2 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a detector match to inspect in context. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.

Reading the findings table: The Severity column indicates the configured finding weight (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — legacy comment-context matches receive a ×1.5 score weight. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier under the legacy scoring model.

Modern AI Meta-Vocabulary12 hits · 30 pts

Severity	File	Line	Snippet	Context
MEDIUM	years/2025.md	394	\| 5) Agentic Misalignment - This study introduces the concept of agentic misalignment, where LLM-based agents autono	CODE
MEDIUM⚡	years/2026.md	29	\| 2) LongHarness - Agents that reason over document collections larger than their context window usually keep retrie	CODE
MEDIUM⚡	years/2026.md	33	\| 6) ReHarness - Agent memory harnesses usually only write forward, so a lesson goes into the bank and is never chec	CODE
MEDIUM⚡	years/2026.md	34	\| 7) The Agentic Scientist - Fully autonomous research agents tend to have a narrow range of ideas and hill-climb my	CODE
MEDIUM⚡	years/2026.md	45	\| 2) CorpusMap - Agents that search large document collections usually see the corpus as a flat set of files, so a r	CODE
MEDIUM⚡	years/2026.md	46	\| 3) VeriHarness - A common practice with long-horizon agents is to sample several rollouts and trust the answers th	CODE
MEDIUM⚡	years/2026.md	50	\| 7) VERA - Long-horizon agents need environments that can be resumed at any stage and scored from observable eviden	CODE
MEDIUM	years/2026.md	66	\| 7) Jev-Mem - Many agentic memory systems use an autoregressive LLM to decide how memories are organized, retrieved	CODE
MEDIUM	years/2026.md	79	\| 4) WFM - More agents now store long-term memory as an LLM Wiki, a folder of markdown pages linked to each other. E	CODE
MEDIUM	years/2026.md	98	\| 7) Salesforce Koa - Custom enterprise models usually need a training set someone has to build. Salesforce trained	CODE
MEDIUM	years/2026.md	101	\| 10) Skill-Based Agentic Evaluation - Storing a fixed reference answer for every eval case breaks when the underlyi	CODE
MEDIUM	years/2026.md	533	\| 2) Codified Context - Single-file AGENTS.md manifests don’t scale beyond modest codebases. A 1,000-line prototype	CODE

Synthetic Comment Markers4 hits · 20 pts

Severity	File	Line	Snippet	Context
HIGH	years/2025.md	689	\| 6) Hallucinations Improve LLMs in Drug Discovery - Claims that LLMs can achieve better performance in drug discove	CODE
HIGH	years/2024.md	308	\| 3) Improved RAG with Self-Reasoning - presents an end-to-end self-reasoning framework to improve the reliability a	CODE
HIGH	years/2024.md	363	\| 2) CriticGPT - a new model based on GPT-4 to help write critiques for responses generated by ChatGPT; trained usin	CODE
HIGH	years/2023.md	555	\| 6) TinyStories - Explores how small LMs can be and still speak coherent English. <br>● Synthetic story dataset: C	CODE

dair-ai/AI-Papers-of-the-Week

Open-Source Code Quality, Security and Technical Debt Review

Risk Review · Payload v1

What These Metrics Mean

Pattern-Signal History

Severity Breakdown

Directory Score Breakdown

Pattern Findings