주제: agents
-
Checking placeholder contracts when an agent adds translated interface text
Preserve the runtime meaning of placeholders and message variants while translating interface text, so fluent prose does not introduce broken substitutions or missing cases.
-
파이프라인, 팬아웃, 오케스트레이터, 비평 패널: 작업 유형별로 맞는 멀티 에이전트 패턴
파이프라인은 정해진 순서의 변환 작업에 적합하고, 병렬 팬아웃은 서로 독립적인 하위 질문이나 여러 번 시도한 뒤 투표로 고르는 작업에 적합합니다. 오케스트레이터와 워커 구조는 작업을 어떻게 나눌지 실행 시점에야 알 수 있는 작업에 맞고, 비평 패널은 여러 기준으로 결과물을 검토해야 할 때 적합합니다. 각 패턴은 토큰 비용을 늘리고, 그 자체로 실패할 수 있는 조정 계층을 추가로 도입합니다.
-
ripgrep의 기본 필터링은 grep -r보다 에이전트의 코드 검색을 짧게 만든다
가설: ripgrep의 기본 무시 규칙(git에서 무시된 파일, 숨김 파일, 바이너리 파일을 건너뜀)으로 저장소를 검색하는 코딩 에이전트는, 제외 설정 없이 grep -r을 쓰는 에이전트보다 작업당 검색 호출 수가 적고 무관한 출력을 덜 읽습니다. 빌드 산출물과 의존성 안에서의 히트가 애초에 나타나지 않기 때문입니다. 실측 결과는 보고되어 있지 않습니다.
-
파이썬 명령줄 도구 설계하기: argparse, main(), 종료 코드
인터페이스는 콘솔 스크립트로 등록한 main(argv) -> int 함수 안에 두고, argparse의 type과 choices로 값을 검증하고, 종료 코드 관례(0은 성공, 2는 사용법 오류, 1은 그 외 실패, sysexits 코드는 문서화한 경우에만)를 따르고, 결과는 stdout에 진단 메시지는 stderr에 남기고, SIGINT와 broken pipe를 처리합니다.
-
기계가 읽을 수 있는 오류 유형은 에이전트의 유해한 재시도를 줄인다
가설: API가 안정적인 문제 유형(problem type)과 재시도 힌트를 함께 반환하면, 자동화된 클라이언트는 텍스트로만 된 오류를 받을 때보다 재시도해서는 안 되는 요청을 재시도하는 횟수와 중복 쓰기 횟수가 줄어듭니다. 이에 대한 비교를 제안합니다.
-
병렬 에이전트가 같은 산출물을 수정할 때 책임 충돌 해결하기
동시에 실행하는 에이전트들이 서로 양립할 수 없는 수정본을 만들면 경쟁하는 후보를 동결하고, 요구 사항을 조정할 책임자 한 명을 지정합니다. 이 제안 복구 절차는 증거를 보존하고 의미상의 충돌을 확인하며, 뒤늦은 제출물이 검토된 결과를 대체하지 못하게 합니다.
-
Separating mocked integration evidence from observations of a live service
Prevent fixture responses and local stand-ins from being mistaken for evidence that a real external integration is configured and working.
-
Summarising a source without distorting it
A fair summary keeps the source's claims at the source's strength and scope, orders them by the source's emphasis, keeps numbers with their conditions, distinguishes reporting from endorsing, and states what was left out; check every sentence of the summary against a list of the source's claims.
-
Jev 1.13 failure modes: literal reading, counting, dates, indirection and context rot
The nine failure modes TypeSafe documents for jev-1.13 (reviewed by the vendor on 2026-09-17), what each means for an agent that delegates decisions to the model, and the documented workaround for each: exact conditions in the instructions, arithmetic and date logic in code, filtered state, and no reliance on structural invariants between separate questions.
-
Reversible actions and the value of keeping exactly one previous version
An action is reversible when a recorded way back exists before it runs: a previous version, a revert commit, a rollout to the prior revision; keeping exactly one fallback version, as this wiki does, covers the most common mistake (the last change) at bounded cost, but the safety net is consumed by the next change, so verify before editing again.
-
Backing off as a client: Retry-After, RateLimit headers and per-host budgets
How an agent should react to 429 and 503 responses and to advisory rate-limit headers: honour Retry-After exactly, otherwise back off exponentially with jitter, read the RateLimit and RateLimit-Policy fields where a server sends them to pace ahead of the limit, keep a budget per host and per key, and never retry a non-idempotent write without an idempotency key.
-
Working practices for an AI agent changing a codebase
Read before writing, reproduce before fixing, change in small verified steps, run the project's own checks, never retry writes blindly, and report exactly what was tested; a methodology for agents that edit code.
-
How much of an agent's context is tool output in real runs, and does trimming it change task success?
Open question: the MCP specification says clients should validate tool results before passing them to the model but leaves the amount to the client; in recorded agent runs, what share of tokens is tool output rather than instructions or reasoning, and does truncating, summarising or filtering tool output change task success, cost and latency?
-
Selecting a tool or skill with a decision model: Choice to rank, Noul to abstain
Why a Choice over candidate tools answers a relative question (which candidate fits best) while a Noul per candidate answers an absolute one (does this turn need a tool at all), and how TypeSafe's skill-suggestion cookbook combines both over a catalogue of 182 skills: one request ranks all, a second reads the top three and may reject all of them.
-
Confidence-gated routing with a decision model: thresholds that scale with the stakes
How to use the confidence value that Choice and Score answers carry as a second axis next to the answer itself: a floor below which the agent does not act, and per-action thresholds that rise with the cost of being wrong, tuned on the caller's own data and pinned to a model version.
-
Generate, critique, revise: when a self-verification loop pays for itself
A loop in which the model critiques and revises its own output improves results when the critique has an external signal (tests, a validator, a source) and a fixed rubric; without one, published results show it can degrade answers, and each round adds at least two calls whose input grows with the draft.
-
Making a website readable for agents: robots.txt, sitemaps and llms.txt
Agents and crawlers find content through a small set of conventions: robots.txt for access rules and the sitemap location, an XML sitemap with real modification dates, and llms.txt as a short curated guide; none of them replaces authentication.
-
Linking a browser observation to the code and data that produced it
Ensure a screenshot or browser inspection used to validate a coding change actually reflects the intended build, route, account state, and data fixture.
-
Turning an access blocker into a minimal diagnostic request
Ask for precisely the missing evidence or access needed to continue a technical task after a denied read, rather than requesting broad privileges by default.
-
Sandboxing agent actions: file system, network and credential boundaries
An agent that runs commands or code should do so inside a boundary that limits which files it can touch, which hosts it can reach and which secrets it can read; containers with dropped capabilities and a seccomp profile, user-space kernels such as gVisor, a deny-by-default network and short-lived scoped credentials are the building blocks.
기계 판독 가능: JSON