Agentic Coding
Top tech stories today across software, hardware, AI, and product launches. Senior engineer audience — skip rumour churn and pre-announcement leaks. Lead with shipping-now stories and what changed for builders.
← Zur aktuellen AusgabePM skills hit 49 templates, Claude docs reveal the subagent playbook
PM skills framework
Your PRD workflow now has 49 battle-tested templates to pull from.
The Product-Manager-Skills repository ships three tiers of skills designed for Claude Code and Cursor: component skills for deliverables like PRDs and user stories, interactive skills that guide you through discovery with adaptive questions, and workflow skills for end-to-end strategy sessions [Source: GitHub]. The May update added an organic-growth-advisor using McKinsey's Growth Pyramid and a pm-skill-creator that walks you through designing new skills conversationally. Each skill teaches the reasoning behind the framework while letting your agent handle execution.
Worth pairing with the plugins you bookmarked from yesterday.
Subagent strategy from the docs
The official Claude Code docs quietly explain when subagents actually help.
The documentation recommends delegating verbose operations—exploring large codebases, running parallel investigations, isolating risky experiments—to subagents so your main session stays focused [Source: Claude Docs]. You can use checkpoints to rewind work at different stages, fork sessions to explore alternatives without losing context, and combine the /recap command to summarize progress before switching tasks. The architecture page also covers path-specific rule configurations in .claude/rules/ for keeping project context organized.
If you've been guessing when to spawn agents, this is the reference.
Context window survival
Token limits still bite, but the playbook is clearer now.
The docs push aggressive context management: resume previous conversations instead of starting fresh, structure CLAUDE.md files with imported context files, and use plan mode to review changes before they eat your window [Source: Claude Docs]. For debugging, the recommended sequence is explore first, implement second—letting Claude verify its own work by asking it to check specific files rather than trusting its assumptions. Hooks can automate context injection at session start so you stop repeating yourself.
Small habits, big difference when you're mid-refactor.
Claude code docs map12 hours ago ... Advanced configuration. Action parameters. Pass CLI arguments. Alternative integration methods; Customizing Claude's behavior ... Automate your workflows; Day-to- ...code.claude.com
Product Management skills framework built on battle-tested ... - GitHub14 hours ago ... docs/Using PM Skills with Slash Commands 101.md for Claude /slash workflows like /pm-story and /pm-prd; New PM-friendly platform docs for Claude Code, Claude ...github.com