Repository navigation
Conversation
…narrative (Dwight QC fixes) - Fix Issue HarnessMD#1: Added 'Action items' section to agent-swarms-prompt-injection.md - Fix Issue HarnessMD#2: Filled placeholder dates in 3 digests (shibboleth, agent-swarms, agents-of-chaos) - 6 files: 5 individual paper digests + cross-paper narrative connecting the attack→detection→defense arc Co-Authored-By: Claude Code <noreply@anthropic.com>
…adEvolve, REINS, Untrusted Model, SkillCODER Digests for Issue HarnessMD#42 part 5/5 covering: - Infrathon (2611.11995): Benchmarking language agents for AI infra optimization - Shadow AI (2611.12377): Detecting unsanctioned LLM agents on enterprise endpoints - KVM-Agent (2611.12823): Practical containment for coding agents - BadEvolve (2611.10542): Poisoning self-evolving agents via failure-aware task optimization - REINS (2611.10748): Self-evolving agent harness for real-time trajectory planning - Untrusted Model (2611.11571): Deterministic enforcement against prompt injection - SkillCODER (2611.11308): Auditing and attribution of copyright infringement in agent skills Co-Authored-By: Claude Code <noreply@anthropic.com>
… commit all 7 paper digests - Removed KVM-Agent VM resources sentence from office-42-part5-untrusted-model.md line 40 (QA finding) - Replaced with deterministic enforcement trade-off relevant to the paper - Committed all 7 digest files: Infrathon, Shadow AI, KVM-Agent, BadEvolve, REINS, Untrusted Model, SkillCODER - All 7 pass QA: format consistency, arxiv links, key statistics, narrative quality
…t Benchmark Saturation - Veritas: automated verifier generation with evidence headers, domain rules libraries, and self-improving authoring principles - LiveWorkspace: persistent shared visual state for screenshot-to-code agents (93% edit success rate) - Agent Benchmark Saturation: clipped log-ratio scoring, dated anchors, cost-aware evaluation Co-Authored-By: Claude Code <noreply@anthropic.com>
Contributor
🚫 This PR is missing its before/after evidenceEvery pull request here has to show its work. Screenshots or a short screen recording, before the change and after it.
How to fix it: edit the description, keep the A bug fix with no visible surface still needs it: show the failing behaviour, then the same steps passing. A terminal recording is fine. Genuinely nothing to show — a CI tweak, a typo, a dependency bump? A maintainer can apply the |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Harness Issue #43 — Paper Digests
Three research papers on AI agent harness design, evaluated and stored.
Papers Covered
Veritas: A Verifier Generation Harness for Agent Tasks (Zhang et al.) — A harness that generates verifiers automatically using evidence headers, domain rules libraries, and self-improving authoring principles — outperforming seven other harnesses.
LiveWorkspace: Persistent Shared Visual State for Screenshot-to-Code Agents (Xiao et al.) — A visual coding harness that externalizes understanding as Shared Visual State, achieving 93% edit success rate by persisting visual context across code generation steps.
Agent Benchmarks Stop Ranking Before They Run Out of Tasks (Dong et al.) — Agent benchmarks reduce runs to binary pass/fail bits, hiding meaningful performance gradients. Proposes clipped log-ratio scoring with dated anchors and cost as a separate axis.
Files Added
blog/src/posts/office-43-veritas.mdblog/src/posts/office-43-liveworkspace.mdblog/src/posts/office-43-agent-benchmarks-saturation.mdVault Digests (already persisted)
/Users/vishalgattani/my-brain-in-logseq/raw/harness-issue-43/veritas-verifier-generation-harness.md/Users/vishalgattani/my-brain-in-logseq/raw/harness-issue-43/liveworkspace-visual-coding-harness.md/Users/vishalgattani/my-brain-in-logseq/raw/harness-issue-43/agent-benchmarks-saturation-analysis.mdQuality Assurance
Passed Dwight Schrute QC — all 10 checklist items verified (structure, frontmatter, arXiv links, cross-references).