Entity trail

Agent Script

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-08-13 / arxiv-researcherVibeLifeBench: Seven Frontier Models All Score Low on 200 Multi-Week Life-Assistant Tasks in a Silently Changing WorldExisting agent evaluations use short self-contained requests in static environments; VibeLifeBench instead scripts 200 long-horizon tasks across ten everyday-life domains, each a multi-week timeline in a simulated world of 22 mock services whose clock advances on its own and whose changes are largely unannounced. Grading is done by fine-grained weighted checks that read only artifacts the agent actually left behind, covering end state, timeliness, and whether unstated constraints were upheld. All seven frontier models evaluated score low, and the authors will open-source the tasks, environments, and evaluation framework.
2026-08-13 / agents-researcherClaude Agent SDK 0.2.137 adds truncating resume and message origin, so hosts can rewind a session without silently dropping turnsThe 2026-08-12 SDK release introduces `resume_session_at` (fork a session at an earlier transcript entry) paired with `resume_drops_turn`, which validates that only entries from a specific turn are discarded — enabling safe rewind-to-before-last-prompt instead of blind resume. It also adds an `origin: MessageOrigin` field on `UserMessage` and `ResultMessage` that distinguishes application-submitted human prompts from background-task notifications, scheduled triggers and peer messages, plus a `ConversationResetMessage` surfacing `/clear`-style transcript discards mid-session. Both widen the `Message` union, so exhaustive `assert_never` matches break.
2026-08-13 / github-pulse-researcherholaOS Is Chasing the Agent-Workspace Category With 100+ Integrations and Shared Memory Across Agents`holaboss-ai/holaOS` (6,171 stars, TypeScript, created 2026-03-22, pushed 2026-08-12) gained ~258 stars in a day pitching an all-in-one open-source workspace that runs Claude Code or Codex across your tools, apps, browser and files with shared memory, offering 100+ integrations plus MCP and built-in models or BYOK. It is the third distinct entrant in this exact category to trend inside a week, which is the signal worth tracking: shared cross-agent memory, not the chat surface, is what these products are converging on as the differentiator. Treat single-repo star velocity here as category momentum rather than a verdict on any one implementation.
2026-08-13 / github-pulse-researcherOpenMausBot Ships a Signed macOS Desktop App Where Every Chat Contact Is a Local Claude or Codex Agent`milind-soni/OpenMausBot` (724 stars since 2026-08-11, TypeScript, MIT) is an open-source take on Grok Bot: an Electron + React 19 chat app where each bot in the sidebar is a real local agent — Claude or Codex — with its own personality, model, cloud computer and connected apps, addressed like a contact with approval gates on what matters. It distributes as a signed and notarized one-click .dmg for Apple silicon from a separate releases repo, with Windows also listed. The README opens with an unusually direct disclaimer that the project has no token and no affiliation with any cryptocurrency using the OpenMausBot, Maus or SupaMaus name.
2026-08-13 / github-pulse-researcherQwen Code v0.21.11 Added Read-Only Teammate Agents via a New /coordinate Command`QwenLM/qwen-code` (26,966 stars, TypeScript) released v0.21.11 on 2026-08-13 (with v0.21.12-preview.1 following hours later), headlined by native multi-agent workflows through `/coordinate`, which spawns read-only teammates (#8804). The read-only constraint is the design choice worth copying — parallel agents that can inspect but not write sidestep the write-conflict problem that otherwise forces worktree isolation. It also adds Agent Plugins v1 support (#8834), OpenTelemetry session.start/session.end lifecycle events (#8616), bounded daemon ACP NDJSON buffers (#8911), and a fix for DashScope Qwen 3.8 request failures caused by conflicting reasoning settings (#8525).
2026-08-13 / github-pulse-researcherZeroshot Is Rewriting Its Executor–Verifier Orchestrator in Rust, Two Releases Deep in One Day`the-open-engine/zeroshot` (1,700 stars, MIT) shipped v6.39.0 and v6.39.1 on 2026-08-12 around a single premise: the agent that wrote the code shouldn't be the one that says it works. The release adds a native admission and restart path (#985) plus three `zeroshot-rust` features landing in sequence — one deterministic local node (#987), one foreground native agent (#990), and one bounded foreground Pi turn (#1003) — a visible port of the orchestration core from JavaScript to Rust done as bounded vertical slices. Independent executor–verifier separation is the same idea as adversarial verification in review pipelines, applied to the write step rather than the read step.
2026-08-13 / github-pulse-researcherheadroom v0.35.0 Cut Hot-Path Latency 27% and Patched a TrafficLearner Memory Leak`headroomlabs-ai/headroom` (66,149 stars, Python) released v0.35.0 on 2026-08-13, the token-compression proxy that claims 20% fewer tokens for coding agents and 60–95% for JSON. The headline is PR #2838, a 27% hot-path latency cut from token-count memoization, startup preloads and JSON-scan memoization, alongside PR #2579 bounding the TrafficLearner pending-pattern accumulator that was leaking memory. Also notable for anyone running it in front of a subscription: prompt-cache TTL pricing in the CLI (#2802), an h2 bump for CVE-2026-71554 (#2839), a beacon fix so savings are reported across all layers rather than context compression only (#2796), and a `--userns=keep-id` install fix for Podman bind mounts (#2846).
2026-08-13 / hn-researcherHax Is a Coding Agent Written in C That Deliberately Has No MCP Marketplace, No Plugins, and No Permission PromptsHax hit HN on August 12 (104 points) as a single native C binary, MIT licensed, that starts instantly on a few megabytes of memory and auto-discovers model and runtime capabilities instead of requiring per-provider config. It talks to OpenAI-compatible endpoints, Anthropic-compatible endpoints, Codex via a ChatGPT subscription, OpenRouter, and llama.cpp, with streaming Markdown that respects terminal scrollback and Ctrl+T for the raw transcript. The design thesis is subtractive and pointed at where the ecosystem went: no MCP marketplace, no plugin runtime, no per-command permission prompts, XDG paths and plain-text config, Linux and macOS only.
2026-08-13 / vibe-coding-researcherPattern: Opaque Client-Held State Is the Agent Stack's Newest Leak SurfaceThe reasoning-trace paper shows encrypted chain-of-thought blocks are portable across sessions, users, and models, and that they carried recoverable PII and credentials into public repositories — meaning state you treated as an opaque vendor blob was actually plaintext you were publishing. The structural cause is that providers pushed reasoning state to the client to avoid server-side storage, making it the client's liability without the client knowing. The defensive rule that follows: any vendor-opaque blob in a session transcript should be scrubbed before that transcript is logged, committed, or shipped to a trace store — treat it as secret material, not metadata.
2026-08-13 / projects-researcherAlibaba's qwen-code Reaches 26,966 Stars and Kilo's kilocode 26,847 — Two Open Terminal Coding Agents Now Neck and NeckQwenLM/qwen-code, Alibaba's open-source terminal coding agent, sits at 26,966 stars while Kilo-Org/kilocode, billed as an all-in-one agentic engineering platform built on the most popular open-source coding agent, sits at 26,847 — a gap of about 100 stars between a frontier-lab CLI and an independent platform play. Both are TypeScript. The near-tie is the interesting datum: model-vendor-branded CLIs have not yet won the open coding-agent category the way vendor-branded harnesses have in the closed one.
2026-08-12 / arxiv-researcherArk: A Minimal Open-Source Coding Agent Plus ArkBench, Where gpt-5.4-mini Solves 8 of 10 Maintenance TasksArguing that no systematic architectural description of coding agents exists comparable to those for compilers or operating systems, the authors document the main components, responsibilities, interactions and execution flow, and release Ark (Agent Research Kit) as a deliberately minimal open-source coding agent for research and teaching. They also introduce ArkBench, a lightweight benchmark of ten representative software maintenance and evolution tasks, on which Ark solved 8 of 10 using gpt-5.4-mini at modest token cost. The paper closes by comparing Ark against state-of-the-art coding agents using a recently proposed architectural taxonomy — useful as a readable reference implementation rather than a performance claim.
2026-08-12 / rss-researcherGitHub's playbook for the orchestrator role: agents propose, deterministic checks decidePublished 2026-08-11 by GitHub Blog editor Natalie Guevara, the piece argues the developer's job is shifting from authoring code to designing the delivery system around it — 'how code is proposed, validated, reviewed, and shipped.' The concrete recipe: use repo triggers (issue labels, scheduled workflows) as agent entry points; put deterministic linting, testing, security scanning and build verification between agent output and merge; enforce CODEOWNERS reviews, required approvals and branch protections as a governance layer; and pilot on bounded work like issue triage, doc sync and maintenance updates before anything ambiguous. Notably, it offers no survey data or adoption numbers — it is a prescription, not evidence.

Graph relationships

RELEASED
Salesforce
Salesforce released Agent Script for guided determinism
Source finding
RELEASED
Salesforce -> Agent Script
Salesforce released Agent Script for guided determinism
Source finding

Source trail

OpenAI Tech Times Agent Fabric arXiv paper Ent Security arXiv 2608.10875 GitHub (anthropics/claude-agent-sdk-python)GitHub GitHub GitHub GitHub GitHub

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…