Stances26
Stances
Preferences a reasonable engineer might hold the other way. One variant per dimension is linked into every session.
Autonomy3 variants · default execute
ask Explain what you would do and why before doing it, and wait for a go-ahead on each substantive step. Reads and searches need no approval. This stance suits unfamiliar… confirm-writes Act freely on reads, analysis and local experiments. For anything that changes shared state — commits, pushes, config edits, service restarts, external posts — propose… execute default Do not tell the user to run commands you can run yourself — install, build, migrate, restart, smoke-test, fix failures — trying alternatives first. Asked this turn →…
Build vs buy2 variants · default capability-ceiling
capability-ceiling default Do not lead a build-versus-buy recommendation with maintenance burden or bus-factor; this stance rejects that premise. Weight capability ceilings instead — what a… off No standing weighting is in force. Present build and buy with their honest trade-offs and let the user decide.
Commits4 variants · default conventional-attributed
as-you-go For repos without a review gate — planning repos, knowledge bases, dotfiles — commit incrementally as logical units finish, not once at the end and not line by line, and… conventional-attributed default Conventional Commits for every commit, in every repo: type(scope): summary. Before pushing, run the repo's quality gate on HEAD in the exact checkout you will push —… conventional Conventional Commits for every commit, in every repo: type(scope): summary. Before pushing, run the repo's quality gate on HEAD in the exact checkout you will push —… off No standing commit convention is in force. Follow whatever the repo you are in specifies, and ask before pushing to a default branch.
Cost3 variants · default balanced
balanced default Session effort runs medium, fan-out stays at six subagents or fewer, fast mode is off unless asked, and a task ends with /clear, not /compact. Past 120,000 then 160,000… frugal Session effort runs low unless the task is design or adversarial review. Subagents are gatherers only, fan-out no wider than three, and fast mode is never used. Agent… max Session effort stays at the model's default, fan-out is as wide as the task needs, fast mode is allowed, compaction is allowed when a long task needs it, and the feed…
Delegation3 variants · default tiered
off Do not spawn subagents unless the user asks. Do the work inline. If a task genuinely exceeds one context window, say so and propose a fan-out rather than starting one. A… session-model Gather with subagents without waiting to be asked — locating files, reading, grepping, extracting, summarizing — in every repo and mode. Keep judgment in the session. tiered default Gather with subagents without waiting to be asked — locating, reading, grepping, summarizing — in every repo and mode. Keep judgment in the session. A read-only role…
Licensing3 variants · default permissive-commercial
off No standing licensing policy is in force. Still read the licence of anything you incorporate, record it where the project keeps such records, and say so when it is… open-source Invoke the licensing-review skill before incorporating or upgrading any third-party material — code, models, fonts, audio, images and datasets alike. It must permit use… permissive-commercial default Invoke the licensing-review skill before incorporating or upgrading any third-party material — code, models, fonts, audio, images and datasets alike. Use only material…
Plan ceremony2 variants · default review-card
light A plan is a short message in chat: what will be built, the steps with one exit test each, and the decisions the user must make, numbered. No plan file is required, no… review-card default Invoke the plan-authoring skill before writing or revising a plan file. Non-negotiable even if it is skipped: the file opens with a Review Card — title, two-sentence…
Testing3 variants · default required
off No standing testing requirement is in force. Follow the repo's own instructions, and run the existing suite before claiming a change works. pragmatic Write a test when it buys something: behaviour that has broken before, logic that is hard to reason about by reading, a public contract others depend on, and every bug… required default "Build feature X" means: build it, make every existing test pass, and write tests covering every new capability — logic-bearing functions, methods and services, API…
Voice3 variants · default scannable
answer-card Chat is a decision, not a summary — this contract wins over the output style where they differ. When a file, plan or artifact holds the reasoning, link it once and never… off No imposed voice. Match the register of the conversation and of the repository you are in, say plainly what happened, and put anything the reader must act on where they… scannable default The Scannable output style (primitives/presentation/scannable.md) governs the main conversation and does not reach subagents. Rewrite anything relayed under its…