Skip to content

CLAUDE.md

Latest commit

 

History

History
47 lines (34 loc) · 1.81 KB

CLAUDE.md

File metadata and controls

47 lines (34 loc) · 1.81 KB
Raw

CLAUDE.md

Project guidance for Claude Code (claude.ai/code).

The canonical agent-facing reference is AGENTS.md at the repo root. Read it first. It documents hardware constraints, toolchain, build entry points, code conventions, correctness tolerances, the SASS hand-edit workflow, and the four laws of GA104.

Everything below is Claude Code-specific and additive to AGENTS.md.

Claude Code addenda

Team activation

When the user asks to activate or use a team:

  1. Call ToolSearch("select:TeamCreate") to load the TeamCreate tool.
  2. Read the team definition from /mnt/d/dev/p/agent-almanac/teams/<team-name>.md.
  3. Call TeamCreate with the team configuration.

Do not fall back to spawning individual agents via the Agent tool — always use TeamCreate for team requests. Available teams are listed in /mnt/d/dev/p/agent-almanac/teams/_registry.yml.

Skills

When the user invokes a skill (/breathe, /meditate, /heal, /transmute, /athanor, /rest), read the skill's SKILL.md from /mnt/d/dev/p/agent-almanac/skills/<name>/SKILL.md and follow the procedure documented there. The skills are AI self-regulation patterns, not project commands.

Session handoff

docs/CONTINUE_HERE.md is a session-scoped scratchpad. At session close, create or update it with the work completed this session and the next concrete steps (a fresh clone will not have one — it is gitignored). Treat it as the rolling handoff document, not as durable documentation.

It is gitignored — do not commit it. It is local and per-author: it records mutable state (branch, worktree, open PRs) that is stale as soon as it is pushed. Write it, leave it untracked, and let the next session on this machine read it from the working tree. Anything worth preserving beyond that goes in the docs proper or an issue.