Help us improve
Share bugs, ideas, or general feedback.
From core
Review what was learned during this session and surgically update CLAUDE.md and AGENTS.md to keep both current with new conventions, patterns, file paths, and workflow discoveries.
npx claudepluginhub yu-iskw/agent-heroes --plugin coreHow this skill is triggered — by the user, by Claude, or both
Slash command
/core:update-claude-mdThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Keep `CLAUDE.md` and `AGENTS.md` accurate and current by reflecting real discoveries made during each session. This skill is invoked automatically at session end via the `Stop` hook.
Captures conversation outcomes into CLAUDE.md — decisions, preferences, patterns, and conventions established during the session.
Updates CLAUDE.md files with non-obvious conventions and gotchas from staged git changes. Activates on user requests like 'update CLAUDE.md' or significant code modifications.
Creates or updates CLAUDE.md following best practices: WHAT/WHY/HOW structure, 30-300 line limits, progressive disclosure with doc references, pointers over copies, verified commands.
Share bugs, ideas, or general feedback.
Keep CLAUDE.md and AGENTS.md accurate and current by reflecting real discoveries made during each session. This skill is invoked automatically at session end via the Stop hook.
CLAUDE.md is for Claude Code and may include Claude-specific extension points (hooks, skills, agents, MCP). AGENTS.md is tool-agnostic and targets Gemini CLI, Cursor, Codex, and other AI assistants — it should contain only information that applies across tools.
Scan the session history for any of the following signals:
If nothing meaningful was discovered — skip all edits and do not touch either file.
Always read the full current CLAUDE.md and AGENTS.md before making any edits. Never modify content you have not read.
| Discovery | Update CLAUDE.md | Update AGENTS.md |
|---|---|---|
| Claude-specific extension points (hooks schema, skills, agents, MCP) | Yes | No |
| General repo structure, file paths, directory layout | Yes | Yes |
General commands (make lint, make format, make test) | Yes | Yes |
| New plugin component conventions (manifest fields, kebab-case) | Yes | Yes |
| New Makefile targets | Yes | Yes |
| New linter in Trunk config | Yes | Yes |
| New CI workflow | Yes | Yes |
| New prerequisite tool | Yes | Yes |
| New key convention (applies to all tools) | Yes | Yes |
| New key convention (Claude-specific only) | Yes | No |
| Discovery type | Target section |
|---|---|
| New directory or file | Repository Structure (tree) |
| New plugin component | Plugin Component Conventions (CLAUDE.md) / Plugin Components (AGENTS.md) |
| New Makefile target | Linting and Formatting → Makefile Targets |
| New linter | Enabled Linters table |
| New GitHub Actions workflow | CI/CD table |
| New convention or rule | Key Conventions (numbered list) |
| New prerequisite tool | Prerequisites table |
| New agent or skill | Agent Orchestration (CLAUDE.md only) |
| Corrected workflow step | Development Workflows |
AGENTS.md.Session discovery: A new make check target was added that runs make format then make lint.
Edits:
CLAUDE.md: add one row to the Makefile Targets table.AGENTS.md: add one row to the equivalent commands section.No other changes needed.