{
  "schema_version": "1.1",
  "id": "weekly-2026-09-14",
  "slug": "2026-09-14",
  "issue_number": "010",
  "title": "feed7 Weekly #010",
  "description": "Six ways to tighten agent tools, tests, permissions, context, and review before your next session.",
  "published_at": "2026-09-14T00:00:00.000Z",
  "modified_at": "2026-09-13T18:05:31.908Z",
  "url": "https://feed7.dev/weekly/2026-09-14",
  "formats": {
    "html": "https://feed7.dev/weekly/2026-09-14",
    "json": "https://feed7.dev/weekly/2026-09-14.json",
    "markdown": "https://feed7.dev/weekly/2026-09-14.md"
  },
  "selection": {
    "rule": "Six source-backed signals and one distraction to leave out.",
    "mode": "ai",
    "ignore_item_id": "auto-014159943e"
  },
  "items": [
    {
      "schema_version": "1.1",
      "id": "auto-79c903650b",
      "slug": "execcritic-learn-to-test-test-to-improve-for-coding-agen-79c903650b",
      "url": "https://feed7.dev/p/execcritic-learn-to-test-test-to-improve-for-coding-agen-79c903650b",
      "title": "ExecCritic: Learn to Test, Test to Improve for Coding Agents",
      "why_included": "Separate test writing from repair, qualify repository-native tests, and freeze them before the repair agent sees execution feedback.",
      "summary": "ExecCritic separates test writing from repair and freezes qualified tests before code changes. On SWE-bench Verified, weak generated tests hurt resolution, while separately trained roles reached 72.6%.",
      "practical_implication": "Test quality determined whether feedback helped on SWE-bench Verified: the no-test baseline resolved 61.2%, base-agent tests lowered that to 57.3%, and GPT-5.6-sol tests raised it to 65.3%. Builders should isolate test generation from patching and reject tests that fail qualification.",
      "agent_context": "ExecCritic assigns testing and repair to separate Qwen-3.5-35B-A3B agents. A fail-closed harness qualifies and freezes repository-native tests, preventing the Repair agent from changing the target after seeing execution feedback.\n\nTest quality determined whether feedback helped on SWE-bench Verified: the no-test baseline resolved **61.2%**, base-agent tests lowered that to **57.3%**, and GPT-5.6-sol tests raised it to **65.3%**. Builders should isolate test generation from patching and reject tests that fail qualification.\n\nSeparately post-training both Qwen roles produced a **72.6%** resolved rate, 11.4 points above the original no-test baseline without stronger-model or Oracle feedback at evaluation. These are benchmark results; repository-specific costs and generalization are not covered here.",
      "source": {
        "name": "arXiv",
        "url": "https://arxiv.org/abs/2609.09133v1",
        "published_at": "2026-09-08T00:00:00.000Z"
      },
      "source_class": "blog_post",
      "content_type": "Paper",
      "layer": "agent",
      "domains": [
        "coding"
      ],
      "topics": [
        "coding-agents",
        "harness-engineering",
        "agent-reliability"
      ],
      "verification": {
        "status": "needs_review",
        "label": "Needs Review",
        "method": "unverified",
        "verified_at": null
      },
      "uncertainty": [
        "Automatically selected from source material; feed7 has not independently tested the claim."
      ],
      "connected_context": null,
      "lifecycle": "New",
      "published_at": "2026-09-08T00:00:00.000Z",
      "modified_at": "2026-09-08T00:00:00.000Z",
      "supersedes": [],
      "expires_at": null,
      "formats": {
        "html": "https://feed7.dev/p/execcritic-learn-to-test-test-to-improve-for-coding-agen-79c903650b",
        "json": "https://feed7.dev/p/execcritic-learn-to-test-test-to-improve-for-coding-agen-79c903650b.json",
        "markdown": "https://feed7.dev/p/execcritic-learn-to-test-test-to-improve-for-coding-agen-79c903650b.md"
      }
    },
    {
      "schema_version": "1.1",
      "id": "auto-b969d2f159",
      "slug": "build-time-vs-run-time-why-dev-tools-fail-in-production--b969d2f159",
      "url": "https://feed7.dev/p/build-time-vs-run-time-why-dev-tools-fail-in-production--b969d2f159",
      "title": "Build-Time vs. Run-Time: Why Dev Tools Fail in Production — Averi Kitsch & Prerna Kakkar, Google",
      "why_included": "Constrain runtime tools with prepared queries, least privilege, bound identity, and limits enforced below the agent prompt.",
      "summary": "Database tools safe for supervised development can be destructive at runtime. Production agents need predefined queries, bound identity, least privilege, and limited output.",
      "practical_implication": "Treat production tool schemas as security boundaries. Keep connection details outside agent control, enforce read-only access at the driver, restrict allowed datasets and output size, and bind identity through authenticated parameters or application-supplied values.",
      "agent_context": "Google separates flexible build-time database tools from constrained runtime tools. In one demo, an agent responded to an error by deleting a table. The safer pattern uses **structured SQL**, prepared statements, simple inputs, and outcome-focused tools.\n\nTreat production tool schemas as security boundaries. Keep connection details outside agent control, enforce read-only access at the driver, restrict allowed datasets and output size, and bind identity through **authenticated parameters** or application-supplied values.\n\nThese controls reduce confused-deputy attacks and blast radius, but do not make the model trustworthy. Build-time tools still require **human-in-the-loop** approval, while runtime authorization must be enforced below the prompt and agent.",
      "source": {
        "name": "AI Engineer",
        "url": "https://www.youtube.com/watch?v=9R--1tg45Jg",
        "published_at": "2026-09-09T00:00:00.000Z"
      },
      "source_class": "video",
      "content_type": "Video",
      "layer": "agent",
      "domains": [
        "coding",
        "security"
      ],
      "topics": [
        "harness-engineering",
        "tool-use",
        "mcp"
      ],
      "verification": {
        "status": "source_linked",
        "label": "Source Linked",
        "method": "source_feed",
        "verified_at": null
      },
      "uncertainty": [],
      "connected_context": null,
      "lifecycle": "New",
      "published_at": "2026-09-09T00:00:00.000Z",
      "modified_at": "2026-09-09T00:00:00.000Z",
      "supersedes": [],
      "expires_at": null,
      "formats": {
        "html": "https://feed7.dev/p/build-time-vs-run-time-why-dev-tools-fail-in-production--b969d2f159",
        "json": "https://feed7.dev/p/build-time-vs-run-time-why-dev-tools-fail-in-production--b969d2f159.json",
        "markdown": "https://feed7.dev/p/build-time-vs-run-time-why-dev-tools-fail-in-production--b969d2f159.md"
      }
    },
    {
      "schema_version": "1.1",
      "id": "auto-f80e17c22f",
      "slug": "500-skills-zero-fine-tuning-linkedin-s-playbook-for-ai-a-f80e17c22f",
      "url": "https://feed7.dev/p/500-skills-zero-fine-tuning-linkedin-s-playbook-for-ai-a-f80e17c22f",
      "title": "500 Skills, Zero Fine-Tuning: LinkedIn's Playbook for AI Agents — Ajay Prakash, LinkedIn",
      "why_included": "Hide large tool catalogs behind search, schema lookup, and execution so agents load only the tools and playbooks relevant to the task.",
      "summary": "LinkedIn scales a large internal agent catalog through search, schema lookup, and execution rather than exposing every tool at once. Its playbooks add task-specific operating knowledge.",
      "practical_implication": "Use progressive discovery when your tool surface grows. Keep playbooks narrow and composable, load their instructions only when relevant, and let agents propose reviewed updates when real work reveals stale or missing guidance.",
      "agent_context": "LinkedIn’s internal system serves more than **1,300 tools** and **600 playbooks** to coding agents. Because performance reportedly degrades beyond 30–40 exposed MCP tools, the catalog sits behind **three meta-tools**: search, schema lookup, and execution.\n\nUse progressive discovery when your tool surface grows. Keep playbooks narrow and composable, load their instructions only when relevant, and let agents propose reviewed updates when real work reveals stale or missing guidance.\n\nThe reported scale includes **8,000 daily users** inside LinkedIn’s centrally managed environment. Its tool-limit observation is not presented as a controlled benchmark, and smaller teams may not need the authentication, telemetry, and distribution infrastructure described.",
      "source": {
        "name": "AI Engineer",
        "url": "https://www.youtube.com/watch?v=9wZpvF3QleU",
        "published_at": "2026-09-09T00:00:00.000Z"
      },
      "source_class": "video",
      "content_type": "Video",
      "layer": "agent",
      "domains": [
        "coding"
      ],
      "topics": [
        "harness-engineering",
        "skills",
        "tool-use"
      ],
      "verification": {
        "status": "source_linked",
        "label": "Source Linked",
        "method": "source_feed",
        "verified_at": null
      },
      "uncertainty": [],
      "connected_context": null,
      "lifecycle": "New",
      "published_at": "2026-09-09T00:00:00.000Z",
      "modified_at": "2026-09-09T00:00:00.000Z",
      "supersedes": [],
      "expires_at": null,
      "formats": {
        "html": "https://feed7.dev/p/500-skills-zero-fine-tuning-linkedin-s-playbook-for-ai-a-f80e17c22f",
        "json": "https://feed7.dev/p/500-skills-zero-fine-tuning-linkedin-s-playbook-for-ai-a-f80e17c22f.json",
        "markdown": "https://feed7.dev/p/500-skills-zero-fine-tuning-linkedin-s-playbook-for-ai-a-f80e17c22f.md"
      }
    },
    {
      "schema_version": "1.1",
      "id": "auto-f7cfdfd7bc",
      "slug": "alibaba-open-code-review-f7cfdfd7bc",
      "url": "https://feed7.dev/p/alibaba-open-code-review-f7cfdfd7bc",
      "title": "alibaba/open-code-review",
      "why_included": "Combine deterministic diff handling with targeted agent review to improve comment placement and precision while preserving broader tests.",
      "summary": "Alibaba’s open-source review CLI combines deterministic diff handling with an LLM agent to improve line-level precision and reduce token use, while explicitly trading away recall.",
      "practical_implication": "Use it when generic coding agents produce noisy findings, skip files, or attach comments to the wrong lines. The deterministic layer selects and bundles files, matches rules, and positions comments; isolated subagents handle related bundles concurrently.",
      "agent_context": "Open Code Review reads diffs, gives an LLM agent code-search and file-reading tools, and returns line-level findings. Its benchmark covers **200 pull requests**, **10 languages**, and **1,505 annotated issues** reviewed by more than 80 senior engineers.\n\nUse it when generic coding agents produce noisy findings, skip files, or attach comments to the wrong lines. The deterministic layer selects and bundles files, matches rules, and positions comments; isolated subagents handle related bundles concurrently.\n\nThe project reports roughly **one-ninth the token use** of a general-purpose agent with the same model, plus higher precision and F1. Recall is lower by design, so it should complement rather than replace broader testing and review.",
      "source": {
        "name": "GitHub",
        "url": "https://github.com/alibaba/open-code-review",
        "published_at": "2026-09-13T00:00:00.000Z"
      },
      "source_class": "tool",
      "content_type": "GitHub Repo",
      "layer": "tools",
      "domains": [
        "coding",
        "security"
      ],
      "topics": [
        "coding-agents",
        "agent-reliability",
        "tool-use"
      ],
      "verification": {
        "status": "needs_review",
        "label": "Needs Review",
        "method": "unverified",
        "verified_at": null
      },
      "uncertainty": [
        "Automatically selected from source material; feed7 has not independently tested the claim."
      ],
      "connected_context": null,
      "lifecycle": "New",
      "published_at": "2026-09-13T00:00:00.000Z",
      "modified_at": "2026-09-13T00:00:00.000Z",
      "supersedes": [],
      "expires_at": null,
      "formats": {
        "html": "https://feed7.dev/p/alibaba-open-code-review-f7cfdfd7bc",
        "json": "https://feed7.dev/p/alibaba-open-code-review-f7cfdfd7bc.json",
        "markdown": "https://feed7.dev/p/alibaba-open-code-review-f7cfdfd7bc.md"
      }
    },
    {
      "schema_version": "1.1",
      "id": "auto-6dde567f1f",
      "slug": "github-copilot-is-now-available-in-the-ai-sdk-harness-la-6dde567f1f",
      "url": "https://feed7.dev/p/github-copilot-is-now-available-in-the-ai-sdk-harness-la-6dde567f1f",
      "title": "GitHub Copilot is now available in the AI SDK harness layer",
      "why_included": "A shared HarnessAgent adapter lets builders compare Copilot with other coding harnesses without rewriting application integration code.",
      "summary": "AI SDK now runs GitHub Copilot behind the same HarnessAgent interface as nine other coding harnesses, using an official adapter and ACP connection.",
      "practical_implication": "If your application already targets HarnessAgent, use the adapter boundary to test Copilot without rewriting application-level agent integration code. Keep agent-specific behavior behind that shared interface.",
      "agent_context": "The AI SDK harness layer adds GitHub Copilot through **@ai-sdk/harness-github-copilot**. It connects to HarnessAgent through **Agent Client Protocol (ACP)** and joins nine other named harnesses, including Claude Code, Codex, and Cursor.\n\nIf your application already targets HarnessAgent, use the adapter boundary to test Copilot without rewriting application-level agent integration code. Keep agent-specific behavior behind that shared interface.\n\nA common interface reduces switching cost but does not establish behavioral parity. The material gives no compatibility matrix, feature differences, or migration constraints across the supported harnesses.",
      "source": {
        "name": "Vercel",
        "url": "https://vercel.com/changelog/github-copilot-ai-sdk-harness-adapter",
        "published_at": "2026-09-10T00:00:00.000Z"
      },
      "source_class": "blog_post",
      "content_type": "Engineering Post",
      "layer": "tools",
      "domains": [
        "coding"
      ],
      "topics": [
        "coding-agents",
        "harness-engineering",
        "agent-sdks"
      ],
      "verification": {
        "status": "official_source",
        "label": "Official Source",
        "method": "source_feed",
        "verified_at": null
      },
      "uncertainty": [],
      "connected_context": null,
      "lifecycle": "New",
      "published_at": "2026-09-10T00:00:00.000Z",
      "modified_at": "2026-09-10T00:00:00.000Z",
      "supersedes": [],
      "expires_at": null,
      "formats": {
        "html": "https://feed7.dev/p/github-copilot-is-now-available-in-the-ai-sdk-harness-la-6dde567f1f",
        "json": "https://feed7.dev/p/github-copilot-is-now-available-in-the-ai-sdk-harness-la-6dde567f1f.json",
        "markdown": "https://feed7.dev/p/github-copilot-is-now-available-in-the-ai-sdk-harness-la-6dde567f1f.md"
      }
    },
    {
      "schema_version": "1.1",
      "id": "auto-0ef977a876",
      "slug": "tech-leads-club-agent-skills-0ef977a876",
      "url": "https://feed7.dev/p/tech-leads-club-agent-skills-0ef977a876",
      "title": "tech-leads-club/agent-skills",
      "why_included": "Treat reusable skills as supply-chain inputs by inspecting instructions, pinning content hashes, and auditing installs and updates.",
      "summary": "A cross-agent skill registry adds scanning, integrity checks, and auditable installs for teams that want reusable coding-agent workflows without blindly trusting marketplace packages.",
      "practical_implication": "Treat agent skills as executable supply-chain inputs. Pin them through lockfiles and content hashing, prefer project-local copies when appropriate, inspect prompts and references, and use the audit log when updating or removing installed capabilities.",
      "agent_context": "Agent Skills packages workflows for Cursor, Claude Code, Copilot, and other agents behind a CLI and MCP server. The repository says marketplace research found **13.4%** of skills had critical issues; its catalog is **100% open source**, contains no binaries, and is scanned before publishing.\n\nTreat agent skills as executable supply-chain inputs. Pin them through lockfiles and **content hashing**, prefer project-local copies when appropriate, inspect prompts and references, and use the audit log when updating or removing installed capabilities.\n\nThe security posture is described by the maintainers, and the marketplace comparison is not substantiated in the supplied material. Human curation and **Snyk Agent Scan** reduce known risks but do not prove that a skill's instructions are safe or suitable for your repository.",
      "source": {
        "name": "GitHub",
        "url": "https://github.com/tech-leads-club/agent-skills",
        "published_at": "2026-09-13T00:00:00.000Z"
      },
      "source_class": "tool",
      "content_type": "GitHub Repo",
      "layer": "agent",
      "domains": [
        "coding",
        "security"
      ],
      "topics": [
        "skills",
        "harness-engineering",
        "agent-reliability"
      ],
      "verification": {
        "status": "needs_review",
        "label": "Needs Review",
        "method": "unverified",
        "verified_at": null
      },
      "uncertainty": [
        "Automatically selected from source material; feed7 has not independently tested the claim."
      ],
      "connected_context": null,
      "lifecycle": "New",
      "published_at": "2026-09-13T00:00:00.000Z",
      "modified_at": "2026-09-13T00:00:00.000Z",
      "supersedes": [],
      "expires_at": null,
      "formats": {
        "html": "https://feed7.dev/p/tech-leads-club-agent-skills-0ef977a876",
        "json": "https://feed7.dev/p/tech-leads-club-agent-skills-0ef977a876.json",
        "markdown": "https://feed7.dev/p/tech-leads-club-agent-skills-0ef977a876.md"
      }
    },
    {
      "schema_version": "1.1",
      "id": "auto-014159943e",
      "slug": "how-a-researcher-uses-codex-and-chatgpt-to-search-for-ne-014159943e",
      "url": "https://feed7.dev/p/how-a-researcher-uses-codex-and-chatgpt-to-search-for-ne-014159943e",
      "title": "How a researcher uses Codex and ChatGPT to search for new antimicrobial molecules",
      "why_included": "The use case lacks workflow, dataset, evaluation, and validation details, leaving no reproducible practice for the next agent session.",
      "summary": "A concrete example of Codex and ChatGPT helping researchers search genomic data for antimicrobial candidates, though the material gives no workflow or validation details.",
      "practical_implication": "For builders, this is a useful pattern for research agents: connect model-assisted exploration to a bounded scientific search space and treat the output as candidates for further evaluation.",
      "agent_context": "César de la Fuente’s lab uses **Codex and ChatGPT** to search **living and extinct genomes** for molecules that might work as antimicrobials against drug-resistant infections.\n\nFor builders, this is a useful pattern for research agents: connect model-assisted exploration to a bounded scientific search space and treat the output as candidates for further evaluation.\n\nThe material does not describe the agent workflow, datasets, candidate count, evaluation method, or experimental results, so it supports a use case rather than a reproducible approach.",
      "source": {
        "name": "OpenAI",
        "url": "https://openai.com/index/using-codex-chatgpt-to-search-for-new-antimicrobials",
        "published_at": "2026-09-10T00:00:00.000Z"
      },
      "source_class": "blog_post",
      "content_type": "Official Release",
      "layer": "industry",
      "domains": [
        "research",
        "data"
      ],
      "topics": [
        "adoption"
      ],
      "verification": {
        "status": "official_source",
        "label": "Official Source",
        "method": "source_feed",
        "verified_at": null
      },
      "uncertainty": [],
      "connected_context": null,
      "lifecycle": "New",
      "published_at": "2026-09-10T00:00:00.000Z",
      "modified_at": "2026-09-10T00:00:00.000Z",
      "supersedes": [],
      "expires_at": null,
      "formats": {
        "html": "https://feed7.dev/p/how-a-researcher-uses-codex-and-chatgpt-to-search-for-ne-014159943e",
        "json": "https://feed7.dev/p/how-a-researcher-uses-codex-and-chatgpt-to-search-for-ne-014159943e.json",
        "markdown": "https://feed7.dev/p/how-a-researcher-uses-codex-and-chatgpt-to-search-for-ne-014159943e.md"
      }
    }
  ],
  "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."
}