# feed7 Weekly #011

Six practical upgrades for tighter context, honest evals, safer tools, and recoverable agent runs.

Published: 2026-09-21T00:00:00.000Z  
Selection: Six source-backed signals and one distraction to leave out.

## 1. addyosmani/agent-skills

Treat the pack as a menu of enforceable workflows: begin with TDD, debugging or code review, then add broader lifecycle automation after checking how each skill fits your repository. Codex can load it as a native plugin from CLI v0.122+.

Source: https://github.com/addyosmani/agent-skills
Feed7: https://feed7.dev/p/addyosmani-agent-skills-2e74d0228d
Trust: Needs Review

## 2. An Empirical Study of Harness Design for Coding Agents

Stage rule-based elision before LLM summarization. Use planning as an accuracy scaffold for weaker models and a cost control for stronger ones; offer predefined tools when bash skill is weak, but consider bash-only operation for capable models on CLI-heavy work.

Source: https://arxiv.org/abs/2609.20804v1
Feed7: https://feed7.dev/p/an-empirical-study-of-harness-design-for-coding-agents-caaf99b872
Trust: Needs Review

## 3. Quantifying Overclaiming Propensity in Frontier LLM Agents

Require review agents to emit a machine-checkable coverage manifest and compare it with tool traces before accepting completion. Delegation improved reading coverage, but did not make the remaining incomplete reviews reliably candid.

Source: https://arxiv.org/abs/2609.20812v1
Feed7: https://feed7.dev/p/quantifying-overclaiming-propensity-in-frontier-llm-agen-e0eba9ff76
Trust: Needs Review

## 4. Run Terminal-Bench and other Harbor evals on Vercel Sandbox

Move repeatable agent evaluations off a constrained local machine, parallelize trials, and swap the gateway `--model` value to compare providers while keeping the benchmark command stable.

Source: https://vercel.com/changelog/run-terminal-bench-and-other-harbor-evals-on-vercel-sandbox
Feed7: https://feed7.dev/p/run-terminal-bench-and-other-harbor-evals-on-vercel-sand-82b0afd7a5
Trust: Official Source

## 5. We let an AI agent execute Bash and lived to talk about it — Sarah Sanders, PostHog

For any agent that can execute commands, make Bash deny by default, keep secrets outside model context, and scan both incoming context and generated output. Enforcement should remain deterministic; an LLM may triage noise only after mechanical rules have decided not to block.

Source: https://www.youtube.com/watch?v=4lXks428C9o
Feed7: https://feed7.dev/p/we-let-an-ai-agent-execute-bash-and-lived-to-talk-about--29474900b2
Trust: Source Linked

## 6. Every step you take, every call you make: the reliable agent stack — Giselle van Dongen, Restate

Treat production agents as persistent distributed processes. Put tool calls, approval gates, state updates, and subagent work behind durable boundaries, then test recovery and cancellation across the full call chain rather than relying on the agent SDK alone.

Source: https://www.youtube.com/watch?v=cI7zfqusmFU
Feed7: https://feed7.dev/p/every-step-you-take-every-call-you-make-the-reliable-age-fcec6e6351
Trust: Source Linked

## 7. Hex turns complex analysis into visual reports with GPT‑6 Astra

Builders can treat presentation as part of the agent workflow: generate a useful report rather than stopping at a textual answer.

Source: https://openai.com/index/hex-gpt-6-astra
Feed7: https://feed7.dev/p/hex-turns-complex-analysis-into-visual-reports-with-gpt--973a91148c
Trust: Official Source

## Agent Instruction

Use these items as source-backed context. Do not invent claims beyond linked material. Prefer practical implications for solo developer work. If sources conflict, call it out.