A study from ETH Zurich and LogicStar (arXiv:2602.11988) found that developer-written AGENTS.md context files improved coding agent success by about 2.4 points in its revised June 2026 version, while …
Claude-mem, an open-source Claude Code plugin at version 13.25.3 with about 94,000 GitHub stars under Apache 2.0, gives Claude Code memory across sessions by capturing lifecycle hooks, compressing the…
Anthropic's Claude Code 2.1.280, documented as of September 23, 2026, treats skills as folders containing a SKILL.md file whose name and description load into context every turn while full instruction…
Anthropic's Claude Code best practices page, checked on 2026-09-23, advises developers to give Claude a runnable check such as tests or a build, plan before multi-file changes, keep CLAUDE.md short, c…
Claude Code v2.1.277 and later reads AGENTS.md by default only when no CLAUDE.md exists in the project, according to a September 23, 2026 comparison of agent instruction files checked against vendor d…
Claude Code does not enforce CLAUDE.md files because Anthropic's memory documentation states the content "is delivered as a user message after the system prompt, not as part of the system prompt itsel…
GitHub Copilot reads `.github/copilot-instructions.md` on every request in a repository, but the file should stay under two pages and avoid task-specific rules, according to GitHub's custom instructio…
Anthropic's .claude directory reference, checked on 2026-09-23 against Claude Code 2.1.280, advises committing CLAUDE.md, .claude/settings.json, .claude/rules/, skills, agents, commands and .mcp.json …
Anthropic's Claude Code memory documentation states that "Each Claude Code session begins with a fresh context window," so corrections typed into chat persist only if written into CLAUDE.md files or a…
A September 23, 2026 blog post outlines a manual workflow for converting accepted, recurring PR review comments into coding-agent rules stored in files such as AGENTS.md, .claude/rules/ and .cursor/ru…
Slack decisions reach AI coding agents only when a person pastes them or the agent searches Slack at the right moment, because Claude Code starts each session with a fresh context window and its auto …
Anthropic removed over 80% of Claude Code's system prompt for models like Claude Opus 5 and Claude Fable 5 with no measurable loss on its coding evaluations, according to a July 24, 2026 post by Thari…
Chroma's July 14, 2025 study "Context Rot: How Increasing Input Tokens Impacts LLM Performance," by Kelly Hong, Anton Troynikov and Jeff Huber, tested 18 models and found that even a single distractor…
Anthropic's Claude Code 2.1.280 documentation and its June 2026 "Steering Claude Code" post lay out a five-mechanism hierarchy for placing team rules, with the guidance that each method "trades contex…
Anthropic advises keeping each CLAUDE.md file under 200 lines, with a hard skip at 4 MiB, because every line loads into context at session start and is resent on every request, the company says longer…
Anthropic's Claude Code hooks, documented in the hooks guide and reference as of Claude Code 2.1.280 on September 23, 2026, run commands deterministically at more than thirty session events, and a Pre…