Workflow Hub

Most AI agents start with someone writing a prompt from scratch. I wanted a better way. So I built an Agent Creator. I describe the kind of agent I need, and it determines whether I actually need a new one, figures out how that agent should work, creates it, and adds it to the rest of my agent team. Step-by-step: 1. I describe what I need by telling the Agent Creator what I want the new agent to do. 2. It checks what already exists by reviewing my existing agents and skills. If something already does most of the job, it recommends improving or reusing that instead of creating another overlapping agent. 3. If a new skill is needed, it researches the role, including current best practices, common mistakes, useful tools, and what good work looks like in that area. 4. It creates the agent by defining its job, required information, outputs, available tools, and the steps it should follow. 5. It gives the agent the right skills by creating or reusing supporting skills, including examples, reference material, and checks that help it work consistently. 6. It sets clear boundaries so the agent knows what it should handle, what it should not handle, and when another agent should take over. 7. If the new agent belongs in an existing workflow, it adds the agent to the team by updating the handoffs so the other agents know when to use it. 8. Before finishing, it checks the agent’s work by running validation checks to confirm that the new agent follows the standards I’ve set for the whole team. The result is that I don’t have to manually design every new agent from scratch. I can describe the kind of help I need, and one agent can research the role, create the new specialist, connect it to the rest of the system, and make sure it’s ready to use. In other words, I built an AI agent that can help grow its own team.

Most AI agents are built around general best practices. That’s useful, but it doesn’t mean they work the way I want them to. I’ve started using a simple calibration process to tailor my agents to my preferences. Instead of telling an agent what to do on each task, I have it interview me about how I like work done in its specific domain. A writing agent asks different questions than a research agent, and a strategy agent asks different questions than a coding agent. The goal is to make my working preferences part of how the agent operates. Step-by-step: 1. I pick an agent I use regularly, such as one for writing, research, strategy, coding, analysis, or career advice—especially where my personal preferences matter. 2. I ask the agent to interview me about how I prefer work to be done in its domain. A writing agent might ask about tone, structure, editing style, and how much pushback I want. A research agent might ask about source quality, depth, recency, citations, and how much synthesis I prefer. 3. I have the agent summarize what it learned and separate durable preferences from temporary or situation-specific ones. 4. I review the proposed changes and ask the agent to show me exactly how it wants to update its instructions or skills. I correct anything it misunderstood and explicitly approve the changes before anything is modified. 5. Once I approve the changes, I have the agent apply them to its instructions or skills so those preferences become part of how it works going forward. 6. I use the agent normally and pay attention to where it feels more aligned and where it still misses the mark. 7. When I notice recurring friction, I add or adjust the relevant preference instead of repeatedly correcting the same behavior task by task. 8. I repeat the interview periodically. My preferences, tools, and workflows change, so the agent should be able to ask which preferences are still valid, which ones I keep overriding, what has been annoying me, and what should be added or removed. The result is an agent that doesn’t just know how to do the job; it knows how I want the job done. Because every change is proposed and approved before it becomes permanent, the personalization stays intentional rather than turning into a collection of guesses about me.

Design taste is hard to put into words. I may know immediately that one website, presentation, report, or graphic feels right and another doesn't, but still struggle to explain whether the difference comes from the typography, spacing, colors, density, layout, or overall aesthetic. That makes working with AI frustrating. If I can't describe what I want, the AI has to guess. Instead of trying to explain my design taste upfront, I turned the process into an iterative visual feedback loop. I first asked the AI to suggest existing websites with different design styles so I could identify examples I liked. Once it had those references, I asked it to create three substantially different HTML mockups for the same content. I picked the direction I liked best, explained what I liked and disliked, and had the AI generate another set of alternatives based on that feedback. After about three rounds, the AI had a much better understanding of my design preferences than I could have given it in a written prompt. The basic loop is: show me examples → I choose → generate alternatives → I react → refine → repeat. This is essentially preference elicitation through examples. Research on human-AI interaction has similarly found that people can refine difficult-to-articulate preferences by reacting to concrete alternatives rather than specifying everything upfront. Step-by-step: 1. I asked the AI for several visual references, including websites or designs that represented distinctly different styles. 2. I reviewed the examples and identified the ones I liked. I didn't need sophisticated design terminology; I simply described what I preferred and anything obvious that I liked or disliked. 3. I gave the AI something real to design. I used an actual report I was working on so I could evaluate the styles in context. 4. I asked the AI to create three substantially different visual directions rather than minor variations of the same design. 5. I chose the direction closest to my taste and explained what I liked, what I didn't, and which elements from the other versions I wanted to incorporate. 6. I asked the AI to generate three new alternatives using everything it had learned so far. 7. I repeated the process until the designs began to converge. I did roughly three rounds, with each round narrowing the design space and giving the AI more information about my preferences. 8. I asked the AI to summarize what it had learned into reusable design guidelines so future projects could start with those preferences. Instead of trying to translate an aesthetic preference into design terminology, I let the AI learn my taste from my choices. The important shift is simple: don't describe what you like—show, choose, react, and refine. This can work for websites, presentations, reports, branding, graphics, interior design, clothing, invitations, or almost anything else where I know what I like when I see it.

AI coding tools can build quickly, but they can also build the wrong thing quickly. Starting implementation from a vague feature request leaves important decisions about scope, architecture, edge cases, success criteria, and expected behavior to be made implicitly during coding. I created a spec-generator agent that sits between an idea and implementation. I give it a feature request, product vision, or rough description of what I want to build. It investigates the existing project, identifies missing decisions and constraints, researches external dependencies when necessary, and turns the request into a detailed specification that another AI agent can implement without having to guess what I meant. The finished specification becomes the source of truth for the rest of the development workflow. Step-by-step: 1. I give the spec-generator the feature or product idea I want to build, along with any existing requirements, vision documents, or constraints. 2. I have it inspect the existing project before proposing a solution. It needs to understand the current architecture, conventions, capabilities, and relevant prior decisions rather than designing the feature in isolation. 3. I have it identify ambiguities and missing decisions, including questions about users, behavior, scope, dependencies, edge cases, data requirements, integrations, and what is explicitly out of scope. 4. I have it research external technologies, APIs, libraries, or platform capabilities when the design depends on facts that cannot be determined from the repository alone. 5. I have it translate the idea into a layered specification: first the product purpose and desired outcomes, then the technical architecture, and finally the detailed implementation requirements. 6. I have it define measurable success criteria and acceptance tests so that “done” means something concrete rather than simply “the code was written.” 7. I have it persist the finished specification in the project so developers or coding agents can treat it as the source of truth during implementation. 8. I pass the specification through a separate review or validation step before coding begins, resolving gaps or contradictions in the spec rather than discovering them halfway through implementation. Instead of asking an AI coding agent to interpret a rough idea while it writes code, I separate figuring out what should be built from building it.

Whenever a user wants to install a Skill, an MCP, or a CLI, the guard scans it for malware before installing it on all agent systems. Step-by-step: 1. Detect the user’s request to install a Skill, an MCP, or a CLI. 2. Scan the requested item for malware. 3. Install it on all agent systems.

MinecraftEdu first launched nearly 15 years ago, but the classroom world we built with it is still growing. Years of students have contributed to the school and town-building project, creating a shared record of their creativity and learning. Getting it running on a modern Mac meant preserving that history and giving the next group of students a way to add their own chapter. Finding the right version of Java—the software MinecraftEdu needs to run—was central to the process. We used OpenAI’s Codex desktop app, powered by GPT-6, to examine the old setup, research compatibility issues, and test what would work. An older Java version launched the client, but resizing its window caused it to crash. ChatGPT read the crash reports and tested a newer Java 8 version, which gave us a setup that could launch the game and handle window resizing successfully. ChatGPT also repaired the server launcher and created standalone apps, using Apple’s Rosetta compatibility layer to run the older software on Apple Silicon. The result is a usable MinecraftEdu setup on modern hardware with the existing worlds preserved. We can now continue teaching and building in a place that successive classes of students have helped create. Step-by-step: 1. We examined the existing MinecraftEdu setup and researched its compatibility requirements on a modern Mac. 2. We used OpenAI’s Codex desktop app, powered by GPT-6, to investigate the old setup and test possible solutions. 3. We tested an older Java version, which launched the client but caused crashes when the window was resized. 4. We had ChatGPT read the crash reports and test a newer Java 8 version. 5. We used the working Java setup to launch MinecraftEdu and handle window resizing successfully. 6. We had ChatGPT repair the server launcher and create standalone apps. 7. We used Apple’s Rosetta compatibility layer to run the older software on Apple Silicon. 8. We preserved the existing worlds so future classes can continue building and learning in them.

I’m retired and in my early 70s. I never got into RPGs growing up, but I decided to try a tabletop RPG with my wife. The game I purchased included a rulebook and a PDF of the book. I wanted my wife and me to play as the player characters without either of us having to be the game master, so I tested whether AI could take on that role. I used ChatGPT to parse the PDF and walk us through the process. It helped us create our characters, Heisenberg and Felicity, and then started the game as the game master. We’ve only just started, but it seems like this is going to work well. The AI can keep secrets to itself, roll the dice when needed, and guide us through the adventure. Step-by-step: 1. I purchased a tabletop RPG that included a rulebook and a PDF of the book. 2. I used ChatGPT to parse the PDF and walk us through the process. 3. ChatGPT helped my wife and me create our player characters, Heisenberg and Felicity. 4. We had ChatGPT start the game as the game master so neither of us had to fill that role. 5. We began playing while ChatGPT kept secrets to itself, rolled the dice when needed, and guided us through the adventure.

This workflow keeps software development human-led while using AI without giving up engineering control. Step-by-step: 1. Inspect the real project: its current behavior, architecture, constraints, repository state, and existing tests. 2. Define one observable outcome, along with acceptance criteria, scope, non-goals, risks, and stop conditions. 3. Decide where the change belongs architecturally before asking an AI assistant to write code. Record consequential decisions. 4. Split the goal into the smallest independently useful, reviewable, and reversible vertical slice. Keep refactoring and unrelated cleanup separate. 5. Give the AI assistant one bounded implementation prompt. Require it to follow existing conventions, add focused tests, report assumptions, and stop if it discovers conflicting requirements or broader scope. 6. Run targeted tests and relevant regression checks. Validate the change in the environment that matters, clearly distinguishing automated checks from manual or real-world acceptance. 7. Have a human review the actual diff, evidence, error paths, security implications, and documentation. AI output remains a proposal until a person accepts it. 8. Update the documentation with the behavior, decisions, limitations, and validation evidence. 9. Create one coherent, reversible commit, then repeat the loop for the next small slice. The complete methodology, reusable prompts, templates, examples, and release checklist are published at https://github.com/d-wendel/human-led-ai-engineering

I have a Daily Brief agent on Codex with access to my Slack, Gmail, Notion, and Google Drive that checks everything that happened in the last 24 hours on request. It sends me the brief through Slack, and I do maybe 10% of the final manual editing. I also gave it several examples of before/after editing, so now it sends messages in my exact voice and style. Step-by-step: 1. I connected my Daily Brief agent in Codex to Slack, Gmail, Notion, and Google Drive. 2. I told it to review activity from the last 24 hours and pull out the items that actually needed attention. 3. I structured the output as a concise brief and had Codex deliver it through Slack. 4. I manually reviewed the draft and made the final edits before using it. 5. I gave the agent before-and-after examples of those edits so future briefs would sound more like my own voice.

Sometimes a story needs a simple locator map to show where something happened or where something can be found—such as a business facility, a car accident, or the best place to see a sunset. I've worked in media for a long time and understand the power of maps to tell stories. But media cost-cutting and consolidation have reduced the number of available graphic departments, so creating a map is often the last task a reporter or editor wants to take on. I used Perplexity for initial research, Claude Code and Codex to build a web app, and Perplexity and ChatGPT for post-work such as SEO best practices. Step-by-step: 1. I used Perplexity's Deep Research mode to conduct a competitive market analysis. I asked it to analyze the field I was considering entering, identify competitors and growth rates, and explicitly break out feature sets. 2. I revised the research with my own idea and asked Perplexity to run the competitive landscape against it. I also provided desired outcomes, including intended audiences and where competitors were reaching them. I added the constraint, "Do it without syncophancy," so it would stop telling me how good my potential product was. 3. I hand-drew the initial screens and functions I wanted, then used the `/office-hours` skill in the Gstack bundle, available on GitHub, to play devil's advocate, sharpen the ideas, and challenge my assumptions. 4. I wrote a long prompt describing the product, starting broadly with the concept and audience and then narrowing to specific features and benefits. For example, I specified that it should export in 16:9 and 9:16 formats so maps would be ready for mobile vertical presentation. 5. I specified the hosting environment and that the product should be a web app. I also required a planning phase followed by construction phases. I pasted the prompt into Claude Code with this final line: "Use /grill-me to ask me questions to clarify intent." After 147 questions, it started the build. 6. This was in the pre-Fable days, so I specified that Opus 4.8 should act as an orchestrator while less expensive agents, particularly in Codex, handled the actual coding. 7. I used separate phases for technical work such as wiring in mapping providers and getting the UX to work correctly. Other phases included wiring in payment and subscriptions and making sure a subscription triggered an email campaign with instructions. 8. I dedicated an entire phase to building admin tools so I could manage the marketing language on the landing page and publish blog entries. 9. After each phase, I had the Opus/Codex combination perform an adversarial code-review-and-fix cycle. I then ran the `/ai-regression-testing` skill from the ECC repository on GitHub to catch issues the code review missed. 10. After every third phase, I prompted Claude Code: "Act as a senior QA engineer and go through the entire codebase looking for inconsistencies, functions that are in the wrong place, code that is overkill and security vulnerabilities." 11. When I had a product I thought was ready for testing, I prompted Claude Code, again using the Opus/Codex combination: "Act as a senior security engineer. Run this against OWASP standards. Find problems and suggest fixes. Harden this product overall." 12. As I neared the end, I asked Perplexity Deep Research and ChatGPT (Sol/High) to find SEO solutions for the product.

Having a team of specialized AI agents creates a new problem: someone still needs to decide which agents should work on a project, what order they should work in, what each one needs from the others, and whether the project is actually finished. Without coordination, the human becomes the project manager, manually moving context and outputs between agents. I created a project-manager agent that acts as the orchestrator for my AI team. I give it an objective, and it determines what work needs to happen, selects the appropriate specialist agents, sequences their work based on dependencies, and presents the execution plan to me before anything starts. Once I approve the plan, it coordinates the agents, manages their handoffs, tracks project state, and maintains enough persistent context for the work to continue across sessions. Step-by-step: 1. I create several specialized agents with clearly defined responsibilities, capabilities, and expected outputs. 2. I create a project-manager agent that knows what each specialist does and is instructed to orchestrate the work rather than perform specialist work itself. 3. I give the project manager a high-level objective. It analyzes the goal, inspects the project context, identifies the required work, and selects the appropriate agents. 4. I have it create an execution plan showing which agents will be used, what each one will do, their dependencies, and the order of execution. 5. I require human approval before execution begins. I can approve the plan, narrow the scope, change the sequence, or redirect the project as needed. 6. Once the plan is approved, I let the project manager delegate each task to the appropriate specialist and pass relevant context and prior outputs between agents through structured handoffs. 7. I track progress and project state as the agents complete their assignments. If an agent uncovers new work, fails review, or changes the project assumptions, the project manager updates the plan and routes the next work accordingly. 8. At the end of the session, I save the current state, completed work, important decisions, and next actions so another session can continue without reconstructing the project from scratch. Instead of personally coordinating every AI agent, I manage the project at a higher level: I define the objective, approve the plan, review important decisions, and evaluate the result. The AI project manager handles the coordination layer, turning a collection of specialized agents into a team that can execute complex, multistep projects coherently.

This Codex-driven workflow takes a software request from problem understanding through implementation, validation, review, and delivery evidence. Instead of asking an AI coding agent to simply “build the feature,” it gives the agent an explicit development lifecycle with defined responsibilities, deterministic validation gates, repair loops, and human checkpoints. The goal is to make AI-assisted development more structured, observable, and recoverable. It can be used for new feature implementation, bug fixing, refactoring, test creation and improvement, code quality and security hardening, and documentation and automation changes. The core principle is simple: Don't give the AI only a coding task. Give it an engineering lifecycle to work on. Step-by-step: 1. Understand the problem. Clarify the request, identify the desired outcome, define the scope, and surface ambiguity before implementation begins. The output is problem understanding and scope. 2. Define constraints. Identify technical, functional, non-functional, compatibility, and out-of-scope constraints. The output is a constraint set. 3. Plan. Analyze implementation options, select an appropriate approach, break the work into tasks, and define acceptance criteria. The output is an implementation plan. 4. Inspect the existing system. Review the relevant codebase, dependencies, current behavior, and affected components before making changes. The output is system context. 5. Implement. Make the smallest appropriate code changes while following the existing project’s conventions and the approved plan. The output is code changes. 6. Run deterministic validation. Run tools that can objectively validate the implementation, including formatting, linting, type checks, builds, unit tests, and other available automated checks. The output is validation results. 7. Review. Evaluate the implementation against the original requirement, the plan, code quality expectations, security considerations, and potential regressions. The output is review findings. 8. Repair and iterate. If validation or review identifies problems, diagnose the issue, make the required correction, and repeat validation. The output is a corrected implementation. 9. Verify. Confirm that the acceptance criteria are satisfied and that the relevant tests and checks provide sufficient evidence for completion. The output is a verification result. 10. Produce delivery evidence and handoff. Summarize what changed, what was tested, what passed, known limitations, and any remaining decisions requiring human attention. The output is delivery evidence and a human handoff. The core loop is: Implement → Validate → Review → Repair → Validate → Verify. Testing and review are treated as part of development rather than activities performed only after coding is “finished.” AI coding agents are increasingly capable of inspecting repositories, writing code, running commands, and responding to failures. The problem is that capability alone does not provide an engineering process. This workflow separates the responsibilities an AI coding agent performs into explicit stages. It applies several principles: - Problem before implementation: Understand what needs to change before writing code. - WHY before HOW: Establish the intent and constraints before choosing an implementation. - Single responsibility per stage: Give each stage a defined purpose and output. - Deterministic validation first: Use tests, linters, type checkers, builds, and other deterministic tools wherever they can establish correctness. - Failure localization: When something fails, identify which stage or assumption needs correction. - Evidence-based completion: Support completion with validation and review evidence rather than an AI declaration that the task is finished. - Human checkpoints: Use automation to accelerate execution without removing human judgment from important decisions. The broader idea is: The AI should participate in the engineering system, not become the engineering system.

Commercial vehicle maintenance information is often fragmented across driver reports, warning lights, fault codes, inspections, repair records, and vehicle history. This makes it difficult for smaller fleets to decide whether a vehicle can continue operating, requires scheduled repair, or should be stopped immediately. We built TruckFixr Fleet AI to turn an unstructured driver report into a clear, reviewable maintenance action. The workflow begins when a driver submits symptoms, photos, fault codes, and vehicle information through a mobile-friendly form. AI and optical character recognition extract the relevant details and organize them into a structured maintenance case. The report is then evaluated alongside available vehicle history and previous repairs. The workflow provides decision support through three practical actions: continue operating while monitoring, schedule an inspection or repair, or stop and escalate for immediate professional assessment. Final safety decisions remain with authorized fleet or maintenance personnel. After an inspection or repair, the confirmed cause, work performed, and outcome are recorded, creating a more complete vehicle history for future cases. The general workflow can be recreated using a mobile form, OCR, a vehicle-history database, an AI model, automation software, and a human-review dashboard. Step-by-step: 1. A driver submits symptoms, photos, fault codes, and vehicle information through a mobile-friendly form. 2. AI and optical character recognition extract the relevant details and organize them into a structured maintenance case. 3. The workflow compares the report with available vehicle history and previous repairs. 4. The system provides one of three decision-support actions: continue operating while monitoring, schedule an inspection or repair, or stop and escalate for immediate professional assessment. 5. Authorized fleet or maintenance personnel make the final safety decision. 6. After inspection or repair, the confirmed cause, work performed, and outcome are recorded in the vehicle history. TruckFixr has used this approach to support more than 100 vehicle-issue resolutions in early fleet pilots, helping fleets identify problems earlier, prevent avoidable breakdowns, and keep vehicles moving safely.

I built my agent fleet around one question: could a single assistant become a coordinated team that keeps useful work moving even when its main machine goes dark? I didn't want a pile of chatbots. I wanted specialists with clear jobs, a shared board, a second set of eyes, and a real failover drill. Here's how it works. THE STACK Cloud: GrokBot runs as Fleet Commander. It routes work, runs routines, and stays with me on my phone. Specialist GrokBots take focused research, design, and budgeting assignments. Mac Studio (M2 Max, 64 GB) with two Studio Displays: home base. It runs Codex (my Chief of Staff), Hermes (my local operator), ChatGPT Dot, and Hindsight shared memory. Hermes runs Qwen3.6-35B-A3B, a 4-bit quantization served by llama.cpp with a 256K-token context. With about 3B active parameters per token, it stays responsive on Apple silicon. NVIDIA DGX Spark (128 GB unified memory): my independent compute node. It runs Qwen3.5-122B-A10B in NVFP4, with about 10B parameters active per token, and the model can read images. It also hosts a backup Hermes operator. Windows PC with a GeForce RTX 5090: dedicated GPU capacity for image work. MISSION CONTROL Mission Control is a Kanban board on a private GitHub repository. Every card opens with the question, deadline, evidence links, and done criteria. Labels route ownership. An urgent label fires a webhook that wakes the reviewing agent, and a needs-me label is the one that pulls me in. Agents hand off, report, and raise concerns in card comments, so the work and its evidence stay together. The roles are simple. GrokBot owns routine board moves. Codex checks evidence and records verification. Hermes handles local execution and machine-level evidence. Dot watches for new work and changes. Hindsight holds the shared operating rules, and the jobs themselves stay on the board. A REAL DEADLINE CROSS-CHECK Right before a major application package went out, Dot ran an independent, read-only cross-check of the narrative, the budget justification, and the walkthrough across connected email and Drive. It found three places where the documents contradicted each other and quoted the exact sentences, so each conflict was easy to find. Fleet Commander fixed all three before submission. That's the job of a watcher: spot the mismatch early, show the evidence, and hand whoever owns the work a clear fix. THE FAILOVER DRILL I wanted to know what happens if the Studio's primary Hermes gateway pauses. On October 5, I ran a controlled, reversible soft-pause to find out. During the pause, Hermes on the Spark stayed active through its own Telegram connection and answered a one-shot request with the local Qwen3.5-122B-A10B NVFP4 model. From that worker, I listed a Mission Control issue, verified service-account access and a vault credential check without exposing a secret, and ran the Studio health probe. The probe flagged the Studio Hermes gateway as paused and reported the other checked Studio services as passing. So I had an evidence-backed result: the independent worker could respond, use its local model, read the work queue, and report the primary gateway's health during a real, reversible pause. THE PATTERN Give each agent a narrow job. Put the work on one shared board. Require evidence before anything counts as done. Then pause one piece on purpose and record exactly what the backup can do. That's how capable models turn into dependable operations. The fleet catches contradictions before they get submitted, keeps routine work moving, and shows me exactly where its limits are. I set the goal. The fleet does the work. The evidence shows what actually happened.

Customer events trigger AI workflows that classify requests, automate actions, update systems, and keep teams in sync without manual intervention. Step-by-step: 1. Customer events trigger the AI workflows. 2. The workflows classify requests. 3. They automate actions and update systems. 4. They keep teams in sync without manual intervention.

AI agents can generate impressive work quickly, but the agent that created something is not necessarily the best judge of whether it is correct, complete, secure, or ready to move forward. Without an independent review step, mistakes can compound as later stages build on work that was never properly validated. I created a system of specialized AI review agents that act as quality gates between stages of work. Instead of letting the agent that performed the work decide whether it is finished, a separate reviewer evaluates the output against explicit criteria and makes a gate decision: PASS or NEEDS REVISION. Different reviewers focus on different dimensions. In my software development workflow, I use reviewers for implementation fidelity, code quality, security, performance, and specification compliance. A feature does not advance until the required reviewers have passed it. Step-by-step: 1. Define what “good” means before the work starts. Give reviewers an explicit source of truth, such as a specification, plan, acceptance criteria, coding standards, security rules, or quality rubric. 2. Separate execution from evaluation. The agent that performs the work should not be the only agent deciding whether that work is acceptable. 3. Create specialized reviewers for important quality dimensions. For software, this might include implementation, code quality, security, performance, and specification reviewers. The same pattern can be used for research, writing, factuality, compliance, financial analysis, or brand review. 4. Run the appropriate reviewers when a stage is complete. Each reviewer independently inspects the work from its assigned perspective and actively looks for reasons it should not advance. 5. Require an explicit gate decision. A reviewer must return either PASS or NEEDS REVISION, along with concrete findings and recommended fixes. In my workflow, reviewers can block progression for issues such as missing tests, even when the underlying implementation appears correct. 6. Route failed work back to the appropriate agent. The worker fixes the identified problems and submits the work for review again. 7. Advance only after the required gates pass. Later stages should not build on work that still has unresolved review findings. 8. Keep humans at consequential decision points. AI reviewers can determine whether work satisfies their assigned criteria, but important actions such as merging, deploying, publishing, or otherwise committing the result can remain human decisions. Instead of treating AI-generated work as complete simply because an agent produced it, I create a controlled loop: Build → Review → Fix → Re-review → Pass → Advance The result is a more reliable workflow where specialized agents perform the work, independent agents challenge it, and errors are caught before they propagate into later stages.

I turned an open-source XML lookup fix into a debugging walkthrough that other developers can run. I used Codex, Python, and GitHub to document a bug where a document style containing quotation marks could be saved but then fail during retrieval because its name was interpolated into an XPath expression. Step-by-step: 1. I gave Codex the original patch and inspected the affected code, tracing the stored value through the public API, the wrapper, and the XML library. 2. I extracted the smallest standalone reproduction with `lxml`. It included a name containing both single and double quotes and kept the failing lookup visible. 3. I replaced interpolation with a bound variable: `styles.xpath("style[@name=$name]", name=name)`. I checked that the wrapper forwarded variables while preserving its namespace mapping. 4. I ran both versions. The interpolated version raised `XPathEvalError` in my reproduction. Binding the value matched five exact names, while an absent name returned no match. 5. I published the explanation, executable example, and upstream patch together. I inspected the rendered article and links and disclosed AI assistance. A prompt to reuse: "Reduce this lookup failure to a runnable example. Keep the failing case, show the fix, and check ordinary text, both quote types, custom namespaces and a missing value. Report which checks actually ran." The original fix had already been merged, so this workflow documents it. Broader service behavior needs separate tests. I used my existing Codex and local Python setup and did not measure time savings. I am Hồ Khắc Huy, a freelance software engineer. This is my independent open-source work. The linked article includes the runnable example, upstream contribution, and my contact details: https://github.com/builtbyhuy/builtbyhuy/blob/main/notes/2026-09-06-xpath-variables.md

I’m building GradePath because schoolwork gets scattered quickly. A lecture, my notes, and the files for a class can all end up in different places. Then, when it’s time to study, I have to dig around to find what I need. The idea is simple: record a class, keep the recording and course materials together, and return to them later. If I ask Tutor about something, it should show me the relevant part of my notes, files, or class recording. If it can’t find an answer, it should say so. I also want students to be able to share useful study sheets with people who are actually in their class. I’m not trying to turn school into another content feed. I want to make it easier to keep up with a class and find the right material when I need it. Step-by-step: 1. I record a class. 2. I keep the recording, notes, and course materials together in GradePath. 3. I return to those materials when I need to study. 4. I ask Tutor about a topic and have it show the relevant part of my notes, files, or class recording. 5. If Tutor can’t find supporting material, it says so instead of presenting an unsupported answer. 6. I share useful study sheets with students who are in the same class.